Title Author(s) Department(s) Abstract
1 Examination
of Nonlinearities in the Van Allen Probes Data During Geomagnetic Storms
Lois Keller
Smith, Michael Liemohn
AOSS In the
recently released Van Allen Probes data, signatures indicating
geomagnetic activity were examined closely. In particular, we chose
several different storms throughout 2012-2013 where a large ionized
oxygen flux was seen in conjunction with a significant drop in the
disturbed storm time (DST) index. For these storms, the electric and
magnetic fields in a cartesian coordinate system were analyzed for
coupled nonlinearities. Preliminary results show that during storm
time, the magnetic field in the z direction will drop dramatically,
sometimes reversing direction in extreme situations. The results of
this work give a clearer pictures of the behavior of the inner
magnetosphere during storm times.
3 Integrating
Gaze Tracking into Virtual Reality for Assessing User Experience
Havens, Eric Maslowski, Theodore Hall
Engineering and Computer Science
With the
advancement of technology, immersive human-computer interaction (HCI)
is quickly becoming vital for modernized computer systems—as seen with
the rise in popularity of virtual reality rooms and tile displays.  These systems, for all their state-of-the-art
technology, are still limited in how well they create a sense of
immersion for the user.  Gaze tracking, if
integrated into these systems, can greatly augment user immersion and
data analysis, and, in doing so, open doors for HCI. 
Once recorded, the user’s point of gaze or pupil diameter
could be utilized to add a new dimension to data collection.  Additionally, real-time gaze tracking could
improve the motion capture system, allowing the virtual environment to
react to where the user is looking rather than just where the user is
This project’s aim is to develop and incorporate an inexpensive gaze
tracking system into the Jugular software system.  This
can be used for desktops/tiled displays, and fully immersive virtual
reality systems such as the Michigan Immersive Digital Experience Nexus
(MIDEN) to enhance future users’ immersive experience and gather new
insight for research purposes.  The Sony PS
Eye camera was modified into an infrared camera. This was accomplished
through low-cost, easy to acquire materials, designing a 3D model of
the lens mount, and rapid prototyping technology.  This
IR camera will be mounted onto the MIDEN’s 3D glasses and used to
actively track the user’s gaze.  The
2-dimensional coordinates will be adjusted to 3-dimensional with
respect to the motion capture data.
4 Discovery
and Design of New Materials via Computer Simulations
Pablo F.
Damasceno, Michael Engel, Carolyn Phillips, Sharon Glotzer
The process
of Self-Assembly – in which the components of a system organize
themselves into an ordered pattern – represents an exciting route for
the design of nano and colloidal materials. Currently, there is no
theory able to predict the structure that will emerge from a system of
locally interacting agents and researchers aiming to isolate the
different factors contributing to this process have to rely on computer
simulations and simplified models. Our work has focused in
understanding the role of some essential factors for self-assembly
(such as shape, Entropy and Enthalpy) as well as using them to
discover, predict and design new target patterns. The utilization of
cutting-edge computer hardware and software, allied to the best
visualization and rendering techniques we have developed, have been
crucial to the progress of this research that can revolutionize
materials and technologies to come.
5 Basic
Social Science on the Web
Jon Atwell Sociology/Center
for the Study of Complex Systems
emergence of new sources of socially-oriented data and the
democratization of research computing power has introduced new
questions and methods to the social sciences. Other than leading to
exciting basic research, these changes offer the opportunity to engage
undergraduate students in new and promising ways. This project does so
through a series of labs in which students gather and analyze these new
data using basic statistics, sentiment analysis, and network analysis.
The labs are built around IPython Notebooks in order to significantly
reduce the skills necessary for students to get and explore these data
while also leaving robust functionality at their finger tips. A trial
run of the labs in an undergraduate seminar has already been hugely
successful. This poster will describe the labs in more detail, present
example results and discuss its effectiveness in the classroom.
6 Mathematical
and Epidemiological Modeling of the Pathogenesis of Human Papilloma
Virus (HPV) in Oropharyngeal Squamous Cell Carcinoma (OSCC)
Dasgupta, Rafael Meza, Thomas Carey, Marisa Eisenberg
Epidemiology OBJECTIVE:
Although cervical and other genital cancers are primarily caused by HPV
infections, recent studies have demonstrated that many individuals also
carry HPV in their oral cavity, and a significant fraction of Head and
Neck Cancers (HNC) contain HPV-DNA, suggesting that the virus may be
the underlying root of these cancers. Indeed, 90% of University of
Michigan (UM) oropharyngeal cancer patients and 50% of nasopharyngeal
cancer patients carry high-risk HPV. Recent research suggests that HPV
is an etiologic agent for a rapidly growing subset of head and neck
squamous cell carcinoma (HNSCC), particularly oropharyngeal squamous
cell carcinomas (OSCC). There remain many open questions about the
potential mechanisms by which HPV induces HNC, as well as the
connection between the rising oral HPV infection and oropharyngeal
cancer incidence. Thus, in order to learn more about the potential
interactions between HPV and HNC, the objectives of this study are to
develop dynamic systems biology models of regulatory networks affected
by HPV and pathways associated with OSCC malignancy and assess the
consequences of HPV infection on the dynamics of tumor cell
proliferation through the use of a sexual behavior and networks
METHODS:  The available genomic,
expression, and physiological data used to develop a mathematical model
of oral HPV natural history of infection was found through a literature
search on the HPV genome, biology of the oropharynx, and HNC. The gene
regulatory networks were mathematically represented with ordinary
differential equations (ODEs), neglecting spatial effects with mRNA and
protein concentrations as continuous time variables. The modeling
software used to analyze and fit these ODEs was Python.
RESULTS: A simple and complex model of the mechanism and transmission
of HPV infection and OSCC carcinogenesis and their corresponding
parameters were proposed and organized from a literature search.
Parameter estimation and fitting of the simple model have produced
preliminary results on HPV infection dynamics, which validates our  methods for further analysis. The draft of the
sexual behavior and networks questionnaire is still being revised.
CONCLUSIONS:  Further analysis, parameter
estimation, parameter fitting, development of genotyping protocols, and
further modeling of the early and late stages of HPV infection in OSCC
are paramount. Moreover, this continuing research will also utilize
information from a sexual behavior and networks questionnaire, to be
piloted in January 2014. The data simulated and collected from the
completion of the two primary aims will result in an overall
HPV-transmission network model with individuals placed in different
OSCC risk groups with different HPV transmission rates, based on sexual
7 Vibrational
Energy Relaxation of Cyanide Ion In Aqueous Solution
Talapatra, Eitan Geva
Chemistry Vibrational
energy relaxation (VER) of molecules in condensed phase provides direct
information about the intermolecular and intramolecular energy exchange
phenomena. For systems like cyanide ion in aqueous solution, the strong
forces generated by the interactions between the ionic species and the
polar solvent molecules, coupled with the quantum effects arising from
the high vibrational energy of the solute (~2080 cm-1), result in an
enhancement of the rate of vibrational relaxation. An in-depth
computational analysis is essential in order to develop a molecular
level understanding of the fast relaxation processes observed in
systems like these.  We have calculated the
VER rate constant for three isotopomers of cyanide ion (12C14N-,
12C15N-, 13C15N-) in H2O and D2O at 300K using Landau Teller formula
where the quantum-mechanical bath force autocorrelation function is
calculated using a linearized semiclassical (LSC) approximation. The
calculations are repeated using classical molecular dynamics
simulations, wherein the quantum effect on the relaxation is accounted
for by using quantum correction factors (QCFs). The calculated VER rate
constants have excellent quantitative agreement with the results
obtained in previous experimental studies. Furthermore, using two
solvent models we have elucidated the plausible relaxation pathways:
(1) A pathway involving the translational and rotational modes of water
as accepting modes (studied using a rigid water model), (2) A pathway
involving the bending modes of water as accepting modes (studied using
a flexible water model). Our results confirm the experimentally
observed trend of relaxation rate constants for the range of cyanide
ion isotopomers. We find that although classical mechanics predicts
that pathway 2 is the dominant one, the stronger quantum enhancement of
the VER rate in pathway 1 compensates for it, thereby making both
pathways viable.
8 A
Computational Approach to Rational Design for Organic Optoelectronic
Phillips, Shaohui Zheng, Zilong Zheng, Francis Devine, Eitan Geva,
Barry Dunietz
Chemistry A
fundamental understanding of charge transfer processes and charge
separation is necessary for efficient development and design of organic
optoelectronic devices. Computational approaches provide fundamental
insights to these processes at the atomistic level, and serve as
cost-effective avenues to guide the synthesis of novel optoelectronic
materials. Conventional density functional theory (DFT) methods, such
as LDA, GGA, hybrid functionals, have been known to fail in accurately
characterizing frontier orbital gaps and charge transfer states in
molecular systems.  In order to address
these failures, we implement the BNL (Baer-Neuhauser-Livshits)
range-separated hybrid (RSH) functional approach within DFT and
time-dependent DFT.
The RSH approach was benchmarked using model systems and then applied
to complex silsesquioxane molecules, which are currently investigated
as candidates for photovoltaic applications. A protocol utilizing
charge-constrained DFT and the polarizable continuum solvent model was
implemented to account for the effect of complex environments on charge
transfer state energies. The protocol was tested and validated by
comparing the charge transfer state energies to experimental
measurements on silsesquioxane molecules. The RSH approach was also
applied to promising molecular systems for pure-organic, phosphorescent
light emitting diode materials, including dibenzophoshole chalcogenide
and bromobenzaldehyde derivatives.
9 A
Treecode-Accelerated Boundary Integral Poisson-Boltzmann Solver for
Electrostatics of Solvated Proteins
Geng, Robert Krasny
Mathematics We present
a treecode-accelerated boundary integral (TABI) solver
for electrostatics of solvated proteins described by the linear
Poisson-Boltzmann equation. The method employs a well-conditioned
boundary integral formulation for the electrostatic potential and its
normal derivative on the molecular surface. The surface is triangulated
and the integral equations are discretized by centroid collocation. The
linear system is solved by GMRES iteration and the matrix-vector
product is carried out by a Cartesian treecode which reduces the cost
from O(N^2) to O(NlogN), where N is the number of faces in the surface
triangulation. The TABI solver is applied to compute the electrostatic
solvation energy of PDB:1A63 (2069 atoms, RNA binding domain of E. coli
rho factor). We present the error, CPU time, and memory usage, and
compare results for the Poisson-Boltzmann and Poisson equations. We
compare TABI results with those obtained using the grid-based APBS
code, and we also present parallel TABI simulations using up to eight
processors. The TABI solver exhibits good serial and parallel
performance combined with relatively simple implementation and
efficient memory usage.
10 High-Fidelity
Aerodynamic Shape Optimization of a Blended-Wing-Body Aircraft
Zhoujie Lyu, Joquim R.R.A. Martins
Aerospace A series of
aerodynamic shape optimization studies for a blended-wing-body aircraft
are performed using Reynolds-averaged Navier–Stokes computational
fluid dynamics with a Spalart–Allmaras turbulence model. A
gradient-based optimization algorithm is used in conjunction with a
discrete adjoint method that computes the derivatives of the flow
equations including the linearization of the turbulence model. A total
of 273 shape and planform design variables are optimized. The objective
function is the drag coefficient at the nominal cruise condition. Lift,
trim, center plane bending moment, and static margin constraints are
considered. The study investigates the impact of the optimization
problem formulation on optimized blended-wing-body designs. Control
surfaces at the rear centerbody are used to trim the aircraft via a
nested free-form deformation volume approach. Trim and static stability
are investigated under both on- and off-design conditions. A comparison
of single-point and multi-point optimization is also presented.
11 Simulation
of Quantum Circuits via Stabilizer Frames
Hector J.
Garcia, Igor L. Markov
Science Engineering
quantum-circuit simulation appears intractable for
conventional computers and may be unnecessary because useful quantum
circuits exhibit significant structure that can be exploited during
simulation. For example, Gottesman and Knill identified an important
subclass, called stabilizer circuits, which can be simulated
efficiently using using group-theory techniques and insights from
quantum physics. Realistic circuits enriched with quantum
error-correcting codes and fault-tolerant procedures are dominated by
stabilizer subcircuits and contain a relatively small number of
non-stabilizer components. Therefore, we develop new data structures
and algorithms that facilitate parallel simulation of such circuits.
Stabilizer frames offer more compact storage than previous approaches
but require more sophisticated bookkeeping. Our implementation, called
Quipu, simulates certain quantum arithmetic circuits (e.g., reversible
ripple-carry adders) in polynomial time and space for equal
superpositions of n-qubits. On such instances, known linear-algebraic
simulation techniques, such as the (state-of-the-art) BDD-based
simulator QuIDDPro, take exponential time. We simulate quantum Fourier
transform and quantum fault-tolerant circuits using Quipu, and the
results demonstrate that our stabilizer-based technique empirically
outperforms QuIDDPro in all cases. While previous high-performance,
structure-aware simulations of quantum circuits were difficult to  parallellize, we demonstrate that Quipu can be
parallelized with a nontrivial computational speedup.
Physics A
clarification on the theoretical predictions of the bias on the
hydrostatic equilibrium mass is increasingly urgent after the claimed
discrepancy on the cosmological parameters derived by the
cosmic-microwave background and cluster counts made by the Planck team.
The literature built on synthetic cluster catalogues suggests bias
values ranging from 10% to 30%, with the largest number justified by
the presence of an additional temperature bias caused by thermal
fluctuations of the medium. In this poster, we make an important step
forward in understanding the different predictions of mass bias by
evaluating the degree of temperature in-homogeneity generated in both
smooth-particle-hydrodynamics (SPH) code and adaptive-mesh-refinement
(AMR) one. On average, mesh-code are more efficient in homogenizing the
medium, especially in the center and in case of no-radiative physics.
In the most external regions, we expect a temperature bias twice as
large in SPH simulations with respect to AMR runs. Once compared with
data obtained from the central regions of ∼ 60 clusters, the level of
temperature structures in AMR are underestimated by about 20% and are
within 10% in SPH for objects with central temperature greater than 5
keV. The inclusion of AGN in SPH simulations improves the agreement to
5% while the introduction of thermal conduction apparently produces an
independence of the temperature dispersion from the cluster tempera-
ture. Constrained on physical mechanisms influencing the cluster
thermal structure will be possible by enlarging the observed sample of
clusters to either higher temperatures or larger radii.
13 Process
Geometric Constraint Mining by Feature-based Analysis
Zhou, Kazuhiro Saitou
With the
increasingly growing demand of lightweight and better crashworthiness
performance on automotive body structural design, the designers
nowadays are more likely to encounter new materials, or even new
manufacturing processes they are not familiar with. The design
guidelines for those new materials as well as processes will be of
great value for designers to improve the
manufacturability/assemblability of the components. Instead of going
through years of trial-and-error practice in manufacturing plant, this
poster proposed a knowledge based process geometric constraint modeling
synthesis through process simulation, design of experiment and data
mining. The extracted process constraints on part geometry are both
interpretable by designers and easy to be embedded into structural
optimization algorithms.
Accelerated Hybrid Volume Surface Integral Equation Solver for
Electromagnetic Analysis of Plasma-Engulfed Vehicles
C. Yucel, Luis J. Gomez, Yang Liu, Hakan Bagci, Eric Michielssen
Engineering and Computer Science
vehicles often are affected by communication blackout upon re-entering
the Earth’s atmosphere. The blackout arises when the vehicle interacts
with the atmosphere around it, giving rise to dense plasmas that are
impenetrable by electromagnetic waves. The vehicle itself often is
covered in a thin and inhomogeneous plasma shell, the density of which
decreases rapidly with distance from the vehicle surface. This plasma
shell hinders the operation of antennas mounted on the side of the
vehicle.  As the vehicle moves through
space, it also leaves behind a large plasma plume. This plume hinders
the operation of antennas mounted on the back of the vehicle. The
nature and density of the plasma shell and wake heavily depend on
operational and environmental conditions and vary rapidly with the
vehicle’s position along its trajectory. To analyze the occurrence of
communication blackout and facilitate the design of robust navigation
systems, fast simulators capable of accurately characterizing the
operation of antennas mounted on plasma-engulfed vehicles are called
The majority of past efforts aimed at analyzing antennas in plasma
environments have relied on finite difference time domain solvers and
were limited to relatively simple antennas and platforms [1].
Unfortunately they do not permit modeling of antennas on realistic
structures. Recently a ray tracing technique was used to analyze
antennas on plasma engulfed re-entry vehicles [2]. Albeit very
powerful, this approach does not allow for a detailed modeling of the
antennas and/or complex plasma structures that may arise in a turbulent
In this study, a hybrid full-wave simulator that addresses the
aforementioned challenges in analyzing scattering and radiation from
plasma-engulfed space vehicles is proposed. The hybrid full-wave solver
uses a surface integral equation solver to model the currents on the
perfect electric conducting surface of the vehicle and a volume
integral equation solver to model the electromagnetic fields in the
plasma surrounding the vehicle. Both solvers were accelerated by a fast
Fourier transform and fast multipole method (FMM-FFT) [3]. The hybrid
solver uses a special block diagonal preconditioner for efficient and
accurate characterization of scattering and radiation from
plasma-engulfed space vehicles in highly inhomogeneous plasma
distributions. Various computational results that demonstrate the
efficiency, accuracy, and modeling versatility of this hybrid full-wave
solver will be presented.
16 Sources of
bias in measures of allele-specific expression derived from RNA-seq
data aligned to a single reference genome
Stevenson, Joseph Coolon, Patricia Wittkopp
of Computational Medicine and Bioinformatics
RNA-seq can
be used to measure allele-specific expression (ASE) by assigning
sequence reads to individual alleles; however, relative ASE is
systematically biased when sequence reads are aligned to a single
reference genome. Aligning sequence reads to both parental genomes can
eliminate this bias, but this approach is not always practical,
especially for non-model organisms. To improve accuracy of ASE measured
using a single reference genome, we identified properties of
differentiating sites responsible for biased measures of relative ASE.
We found that clusters of differentiating sites prevented sequence
reads from an alternate allele from aligning to the reference genome,
causing a bias in relative ASE favoring the reference allele. This bias
increased with greater sequence divergence between alleles. Increasing
the number of mismatches allowed when aligning sequence reads to the
reference genome and restricting analysis to genomic regions with fewer
differentiating sites than the number of mismatches allowed almost
completely eliminated this systematic bias. Accuracy of allelic
abundance was increased further by excluding differentiating sites
within sequence reads that could not be aligned uniquely within the
genome (imperfect mappability) and reads that overlapped one or more
insertions or deletions (indels) between alleles. After aligning
sequence reads to a single reference genome, excluding differentiating
sites with at least as many neighboring differentiating sites as the
number of mismatches allowed, imperfect mappability, and/or an indel(s)
nearby resulted in measures of allelic abundance comparable to those
derived from aligning sequence reads to both parental genomes.
17 Scoring
Protein Interactions using CRAPome – a Contaminant Repository for
Affinity Purification Mass Spectrometry Data
Mellacheruvu, Zachary Wright, Anne-Claude Gingras, Alexey Nesvizhskii
DCMB ffinity
purification coupled with mass spectrometry (AP-MS) is widely used for
the identification of protein-protein interactions. However, in a given
experiment, distinguishing between true interactions and background
contaminants (e.g. proteins interacting with the solid-phase support,
affinity reagent or epitope tag) is a challenging task. While the
standard approach is to use negative controls to identify background
contaminants, most small-scale AP-MS studies fail to comprehensibly
characterize the background. Negative controls, however, are
bait-independent and aggregating data from multiple AP-MS studies can
increase the coverage. Hence, we developed a curated and annotated
repository of negative controls called the Contaminant Repository for
Affinity Purification Mass Spectrometry data (CRAPome). We present here
the repository and methods to score protein interactions using the
18 Uncovering
pH-dependent Excited States in Proteins using pH-based Replica Exchange
Constant pH Molecular Dynamics Simulations and GPU Acceleration
Garrett B.
Goh, Laricheva, E. N., and Charles L. Brooks III
Chemistry pH plays a
crucial role in regulating biological activity such as protein folding,
enzyme catalysis and protein-protein interactions, and understanding
the protein’s dynamical response to pH changes will improve our
fundamental understanding of such processes. Recent advances have
demonstrated the increasing importance of transiently populated minor
states in protein dynamics and function, but due to their low
populations, pH-dependent excited states are difficult to characterize
experimentally. Here, we report on the development and application of a
novel computational method, an explicit solvent constant pH molecular
dynamics (CPHMD) framework to model realistic pH-coupled dynamics in
proteins. In CPHMD, the protonation states of residues are dynamically
changing in response to the external pH environment and conformational
dynamics of the protein. To improve the convergence of our simulations,
we utilize a pH-based replica exchange sampling method, which is a
loosely coupled and highly parallelizable algorithm. In addition, we
utilized new GPU technology that is capable of running molecular
dynamics simulations orders of magnitude faster than traditional CPU
hardware to improve the sampling of our simulations. To probe the
atomistic-level details of pH-dependent excited states, CPHMD
simulations was used to interrogate the dynamics of a series of
staphylococcal nuclease (SNase) mutants with buried ionizable residues
at different pH environments. Amongst the key findings from our
simulations is the existence of open state structure for these “buried”
residues, where local solvation around the protonating site was
observed in all SNase mutants with highly shifted pKa values. In the
most extreme examples of pKa shifts recorded by experiments, these open
state structures are transiently populated at physiological pH,
contributing a small but non-zero population of 1-2%. Nevertheless,
sampling these open states was discovered to be a necessary condition
for accurately reproducing experimental pKa measurements, to which our
calculated pKa values demonstrated good agreement with experiments with
a low average unsigned error of 1.3 pKa units and correlation
coefficient (R2) of 0.78. The work we present here provides the first
validation that buried ionizable residues can readily evoke
pH-triggered conformational fluctuations that are propagated into
physiological pH range as excited states. Furthermore, the good
agreement between calculated and experimental pKa values also
demonstrate that explicit solvent CPHMD simulations are able to model
pH-coupled dynamics in both hydrophilic and hydrophobic environments,
paving the way for future applications of such as modeling pH-dependent
protein folding and biological activity of membrane proteins.
19 Accelerating
the Discrete Element Method using GPUs in HOOMD-Blue
Spellings, Ryan Marson, Sharon Glotzer
faceted shapes are commonly created in experimental systems of
colloidal and nanoscale particles, many interesting physical phenomena,
such as crystalline nucleation and growth and glassy dynamics, are
challenging to model in these systems. A variety of computational
models have been developed over the years, but each has its own set of
advantages and disadvantages.

The Discrete Element Method (DEM) is used extensively to study granular
systems of polyhedral particles, but existing implementations are
serial and limited in performance. We have implemented a parallel
version of the Discrete Element Method (DEM) on the GPU within the
HOOMD-Blue framework. By decomposing the force calculation into its
components, this implementation can take advantage of massive data
parallelism, enabling optimal use of the GPU for even relatively small
systems while achieving a speedup of 65 times over a single CPU core.
Overall, this method is a natural extension of classical molecular
dynamics into the realm of faceted particles, and allows simulation of
size scales ranging from the nanoscale to granular particulates within
the same simulation framework

20 The
Seasonal Migration of H3N2
Zinder, Manojit Roy, Edward B. Baskerville, Mercedes Pascual
DCMB Influenza A
(H3N2) is circulating in the human population since 1968 and continues
to maintain considerable morbidity and mortality. Genetic sequences
sampled since its introduction serve for combined evolutionary and
epidemiological inference. Despite it being highly seasonal, the role
of seasonality in determining global migration and persistence patterns
has not been directly established, in lack of an adequate computational
model, and due to confounding factors such as air-travel and
demographics. Using a Bayesian Metropolis-Coupled MCMC framework, we
explore the seasonal migration of H3N2, integrating in parallel over a
massive number of phylogenetic tree topologies and model
configurations; using open source software developed in our lab and the
University of Michigan’s HPC system. We initially focus on China and
North-America and later on multiple global communities, showing key
windows in time, in which successful invasion takes place. We attribute
H3N2 migration seasonality to high incidence and growth periods at the
source community and matching epidemic troughs at the destination.
Quantifying and improving our understanding of H3N2 migration, will
further our understanding of flu ecology, expand the horizon of
prediction, and eventually lead to better matching vaccines;
identifying candidate strains that are not only able to escape immunity
but also have the right seasonal opportunity to establish themselves.
21 A
Well-Conditioned Volume-Surface Field Integral Equation (VSCFIE) for
Inhomogeneous Cylindrical Scatterers with High-Electrical Contrasts
Luis J.
Gomez*, and Eric Michielssen
EECS Volume
integral equations (VIEs) are commonly used to analyze scattering from
inhomogeneous dielectric objects.  Unfortunately,
when applied to high-contrast objects, their discretization calls for
very fine meshes and results in ill-conditioned systems of equations.  This severely limits the utility of VIEs in
modeling electromagnetic phenomena germane to a host of biological and
geophysical applications.

A partial solution to this problem was proposed by Usner and coworkers,
who preconditioned the VIE using a surface integral equation (SIE).
(Usner et al., IEEE Trans. Antennas Propagat., vol. 54, pp. 68-75,
2006.) First, they subdivide the object into subscatterers each having
a small (controlled) ratio of maximum to minimum permittivity. Next,
they wrap each subscatterer in equivalent electric and magnetic surface
currents and model interactions between subscatterers using SIEs; this
procedure allows for an artificial increase in the effective
permittivity of the “background medium” in which each subscatterer’s
polarization currents radiate. Finally, all surface and volume currents
are determined by numerically solving a system of volume-surface
integral equations (VSIEs) consisting of coupled combined field
integral equations (CFIEs) and VIEs. The VSIE alleviates the need for
very fine meshes. However, if the electrical size of the scatterer is
fixed, the condition number of the matrix resulting from the
discretization of the VSIE grows linearly with the scatterer’s maximum
effective permittivity. Furthermore, if the effective permittivity of
the scatterer is fixed, this matrix’ condition number is inversely
proportional to the scatterer’s electrical size.

Here, we propose a method that alleviates both these problems.  We start from a subdivision of the object into
subscatterers inspired by that used by Usner and coworkers, and
introduce equivalent electric and magnetic currents on their surfaces.  Next, we solve for the surface and volume
currents using a system of Volume-Surface Integral Equations (VSIE)
consisting of coupled discrete CFIEs and new, combined VIEs.  For each subscatterer, a combined VIE is
constructed by judiciously adding contributions due to the currents
exterior to it and propagating in the “background medium”. 
Numerical data obtained by analyzing time-harmonic TE
scattering from various 2D layered cylinders suggests that
discretization of the new VSCFIE yields matrices with condition numbers
that are unaffected by the scatterer’s maximum permittivity and
electrical size.

22 Tuning the
HOMO-LUMO Gap in Conjugated Polymers for Organic Photovoltaics
Applications based on First-Principles Calculations
Xiao Ma,
Hossein Hashemi, Bonggi Kim, Jinsang Kim, John Kieffer
science and engineering
To tune the
HOMO and LUMO energy levels via alternating donor-acceptor monomer
units, we investigated a series of conjugated polymers (CP)s in which
the electron withdrawing power of the acceptor group and the electron
giving power of the donor group is varied, while maintaining the same
conjugated chain conformation. We observed that the introduction of
electron withdrawing groups lowers the LUMO level, while keeping the
HOMO level almost unchanged. Conversely, inserting the electron
donating groups raises the HOMO level while maintaining the LUMO level
unchanged. According to these trends, designing a low band gap polymer
requires strong donors and acceptors.  Using
first-principles calculations we investigated underlying reason. Charge
localization on the electron-rich and electron-poor segments in CPs
plays a key role. We identified strong correlations between the
withdrawing strength of the acceptor group, the HOMO and LUMO levels,
and the degree of orbital localization, which allows us to derive
reliable design principles for CPs.
23 Employing
Machine Learning for Fast, Accurate, and Easily Accessible Predictions
of Activation Energies
Jordan N.
Metz*, Paul M. Zimmerman
Chemistry Fast and
reliable predictions of the favorable products formed from given
reactant species at set conditions are of great interest to aid in the
design of synthetic procedures.  Approaches
towards obtaining this insight have utilized a number of techniques
including quantum chemical transition state search methods, rule-based
expert systems, and machine learning.  Although,
these rule-based expert systems and machine learning techniques bypass
the expense of quantum chemical transition state search methods, the
currently proposed models suffer several drawbacks including complexity
and difficulties in the transfer to additional reaction types among
others.  In our work, we demonstrate a
simple machine learning technique that directly predicts the activation
energy using common quantum properties of reactant-product species
capable of providing insight across a variety of reaction types.  Initial pericyclic studies reveal activation
energies within ~2kcal/mol of quantum chemical calculations on average
where the prediction timescale occurs on the order of milliseconds.
24 Python
Enabled Atomistic Simulation and Analysis of Highly Cross-linked Epoxy
Sebeck, Michael J. Waters, John Kieffer
Science and Engineering
cross-linked glassy polymers such as epoxy are widely used in
industrial adhesive and polymer matrix composite applications. The
mechanical properties of these structures are highly dependent on the
topological nature of the network. Prediction of the mechanical
properties of these materials is vital for efficient design. However,
this fundamental topology of the network is difficult to probe
experimentally. A series of epoxy structures have been generated using
a combination of molecular dynamics simulations in LAMMPS and a
Python-enabled dynamic polymerization code. We examine the effect of
amine functionality and system size on the behavior of these network
25 Kirke: A
Python toolkit for molecular simulations.
M. J.
Waters, K. Sebeck, E. J. Coyle, and J. Kieffer
Science and Engineering
simulation is a field rife with different file formats and bespoke
analysis methods. Even with standard simulation codes such as LAMMPS,
VASP, AMBER, etc., there are many one-time-use codes written for
pre/post-processing of molecular structures. We sought to reduce
redundant coding labor in our simulation group by creating a shared
library for the reading, writing, manipulation, and analysis or
molecular simulation data. By using a common data structure for our
molecular structures, all of these collaboratively written functions
can be used as agnostically as possible of input or output format.
Using this system, we have saved significant intellectual overhead
simply through elimination of redundancy. The library is written in
Python 2.x with some functional dependence on NumPy, SciPy, and
26 High-Throughput
Screening for Carbon Capture and Chemical Energy Storage Materials
Hyun Seung
Koh, Malay Kuma Rana, Jacob Goldsmith, Antek Wong-Foy, Michael
Cafarella, Donald J. Siegel
frameworks (MOFs) have recently emerged as promising materials for
carbon capture and methane storage. Even if there are over thousands of
MOFs were synthesized, and proposed over 100,000 MOFs, there are lack
of knowledge about gas adsorption. Thus, we identify MOFs structures
from Cambridge Structural Database, and clean the structures that we
find. We calculate H2 uptake of 22,800 cleaned MOFs by Chahine rules,
and discover there is a limitation increasing both volumetric and
gravimetric uptake. Also, Grand Canonical Monte Carlo (GCMC) results
show the same relation is in CH4 simulations, and we found some
untested good MOFs for H2, and CH4 storage. Amongst these many possible
MOFs, metal- substituted compounds based on M-DOBDC and M-HKUST-1 have
demonstrated amongst the highest capacities for CO2 and CH4 at moderate
pressures and temperatures.  Here we also
explore the possibility for additional performance tuning by
computationally screening several metal-substituted variants of these
compounds with respect to their CO2 adsorption enthalpies and CH4
capacities. In the case of CO2, our screening identifies 13 compounds
having adsorption enthalpies within the targeted thermodynamic window
-40 to -75 kJ/mol. Also, the partial charge on the coordinatively
unsaturated metal sites is found to correlate with the adsorption
enthalpy, suggesting that this property may be used as a simple
performance descriptor for carbon capture efficiency. 
For methane storage, calculated adsorption enthalpies are
found to be 10-20 kJ/mol less exothermic than for CO2, consistent with
a weaker, dispersion-based CH4—MOF interaction. In parallel with these
thermodynamic analyses, methane adsorption isotherms were predicted
using GCMC simulations across the remainder of the M-DOBDC series, with
additional comparisons to prominent MOFs. PCN-11 & Be-DOBDC yield
the best combination of usable gravimetric and volumetric methane
densities at pressures below ~50 bar, while MOF-5 is best at higher
27 Self-assembly
of colloidal complementary shape alloys
Eric S.
Harper, Ryan L. Marson, Joshua A. Anderson, Sharon C. Glotzer
Science and Engineering
of nanoparticles and colloids holds great promise to create useful new
materials. Nearly ubiquitous in nature, self-assembly is a powerful
tool for organizing matter, observed in everything from the formation
of virus capsids and protein folding to the bubbles on top of a glass
of beer. Scientists and engineers in a wide variety of disciplines hope
to better understand how different particles self-assemble to
facilitate the engineering of desired assemblies of materials. While
the use of DNA tethers, chemical functionalization, and even magnets
have been shown to effectively direct self-assembly, the effect of
particle shape on self-assembly has received surprisingly little
attention. Nature makes extensive use of shape-based complementary
interactions, such as the “lock and key” mechanism of enzymes and
protein recognition at the surface of a cell. Utilizing both
MPI-accelerated and GPU-accelerated Metropolis Monte Carlo simulations,
we investigate a class of complementary “shape alloys” that can be used
to stabilize desired phases. These complementary particles fit together
in a manner similar to that of puzzle pieces. We investigate two
dimensional systems of polygons: squares, pentagons, and hexagons.
These shapes are split into halves as well as complementary halves e.g.
squares into rectangles and right isosceles triangles. By using our
visualization and analysis software, we see evidence that this shape
complementarity promotes the desired assembly, even in a system which
otherwise would not assemble e.g. the right isosceles triangles.
28 Hot carbon
corona in Mars’ upeer thermosphere and exosphere: solar cycle and
seasonal variability
Yuni Lee,
Michael R. Combi, Valeriy Tenishev, Stephen W. Bougher
AOSS This work
presents the variability over seasons (i.e., orbital position) and
solar cycle of the Martian upper atmosphere and hot carbon corona. We
investigate the production and distribution of energetic carbon atoms
and the impacts on the total global hot carbon loss from all dominant
photochemical processes at five different cases: AL (aphelion and solar
low), EL (equinox and solar low), EH (equinox and solar high), PL
(perihelion and solar low), and PH (perihelion and solar high). We
compare our results with published results, but on only limited cases
because of the dearth of studies on solar EUV flux and orbital position
variations. Photodissociation of CO and dissociative recombination of
CO+ are generally regarded as the two most important source reactions
for hot atomic carbon. Of these two, photodissociation of CO is found
to be the dominant source in all cases considered.
To describe self-consistently the exosphere and the upper thermosphere,
a combination of our 3D Direct Simulation Monte Carlo (DSMC) model and
the 3D Mars Thermosphere General Circulation Model (MTGCM) is used. The
basic description of this hot carbon calculation can be found in the
companion paper to this one (Lee et al. 2013a). The spatial
distributions and profiles of density and temperature and atmospheric
loss rates are discussed for the cases considered.
29 Computational
Simulation of Seismic Collapse Capacity of Steel Columns
Fogarty, Sherif El-Tawil
Civil &
Environmental Engineering
The effect
of local buckling due to lateral loading on the axial capacity of steel
columns is investigated using detailed finite element models at the
member, subassemblage, and system levels. An overview of past research
addressing column failure under seismic loading is presented to
characterize the current state of knowledge. Preliminary simulation
results indicate that flange local
buckling due to lateral loading could significantly reduce the axial
resistance of steel moment frame columns.
of steel moment frame columns.
31 Cluster
Stacking: Increasing Precision and Accuracy in Galaxy Cluster Mass
Kern, Daniel Gifford, Christopher Miller
Astronomy Galaxy
clusters, regions in the universe where galaxies are found to be
clustered together, are the largest gravitationally bound structures in
the universe, and play a crucial role in the study of the evolution of
the universe. By measuring the total (dark matter) mass of these
clusters we can analyze how clusters form and can even constrain
fundamental parameters of the evolution of the universe. However,
measuring the total mass of galaxy clusters accurately and precisely is
one of the major barriers in observational cosmology. The caustic
technique is one competing method, which uses the newtonian kinematics,
positions and velocities, of the clustered galaxies to estimate the
mass of the underlying dark matter content (Gifford et al. 2013).
However, the precision and accuracy of the caustic technique suffers
greatly from low number statistics. We show that by taking multiple
clusters with low number statistics, and stacking their phase space
together to create an ensemble cluster, we can greatly reduce the
intrinsic bias and scatter, and obtain precise and accurate mass
estimates of clusters with sample sizes as low as N ~ 5. We use N-body
simulation data to analyze the statistical improvements the stacking
method provides. However, this can only be done if we stack data on
anywhere from 20 ~ 100 clusters, each needing an array of independent
realizations of different N, quickly creating the need for high
performance computing to solve the issue. By running ~2,000 in-parallel
jobs on the FLUX high performance computer, we successfully show that
within the Millennium Simulation, the stacking method applied to the
caustic technique can increase the precision of an ensemble cluster
mass estimate by upwards of 60%, and increase the accuracy by upwards
of 50%.
32 Numerical
simulations of accretion outbursts during low-mass star formation
Jaehan Bae,
Lee Hartmann, Zhaohuan Zhu, Richard Nelson
Astronomy During
their formation, low-mass (< 1 solar mass) protostars are thought to
“episodically” accrete material in large, rapid bursts from their
circumstellar disks. While such events have been observed over the last
several decades, their mechanism(s) by which these accretion bursts
occur is still under debate. In this study, we use 2D Fast Advection in
Rotating Gaseous Objects (FARGO) code to model accretion bursts driven
by magnetorotational instability (MRI), which successfully reproduce
some observed outburst properties.
33 Thermal
Buckling of Composite Plates with Spatially Varying Fiber Orientations
Adam Duran,
Nicholas Fasanella
buckling analysis of square composite laminates with variable stiffness
properties is presented. Fiber angles vary spatially and result in
material properties that are a function of position. The critical
buckling temperature for such laminates were obtained based on
classical lamination theory and the Galerkin finite element method
using Kirchhoff plate elements. In our approach, we discretize the
domain to transform nonlinear fiber path functions to linear piecewise
functions so we may apply classical lamination theory. 
Using this method, thermal responses for symmetric
balanced simply supported laminates under constant thermal load were
investigated and the optimal angles and fiber paths to resist thermal
buckling were obtained. Validation for the method presented is achieved
using the results for the special case of constant fiber angles found
in literature. We found that angle configurations exist that provide
higher resistance to thermal buckling in comparison to straight fiber
34 Target
research areas for High Throughput Nuclear Architecture Analysis (HTNAA)
Allyn-Feuer, David Dilworth, Walter Meixner, Alex Ade, and Brian Athey
DCMB Researchers
at the University of Michigan, in collaboration with the Brady
Institute at Johns Hopkins University, have begun development of a
high-throughput, semi-automated cell analysis system to characterize a
number of cellular features including chromosome territory boundaries
and their three-dimensional arrangement within the cell nucleus.  Image data acquired with confocal and
super-resolution imaging techniques will be processed with an image
processing pipeline facilitating tasks such as channel separation,
interest operations and classification, spectral unmixing, clustering,
data fusion, feature extraction, and visualization across large
populations of cells.  This information,
collected in an appropriate biocellular system with appropriate
probesets, can be used to interrogate biologically and clinically
relevant hypotheses by defining and calculating derived data forms
tailored to the investigation.  Here we
explore four major research areas which will benefit from the
application of HTNAA techniques: Transcription Factory research, the
validation of Hi-C results, Prostate Cancer progressivity prognosis,
and the exploration of Glucocorticoid receptor mediated stress diseases.  We also explore the derived data forms which
will contribute to these investigations.
35 Improving
Quality of Service and Fairness in Stochastic Operating Room Planning
Yan Deng,
Siqian Shen, Brian Denton
IOE Existing
research on stochastic operating room (OR) planning rely on an
expected-cost-based approach, which penalizes under-performance and
minimizes the total expected cost. However, penalty costs are often
difficult to estimate accurately. In this paper, we formulate the
problem as chance-constrained programs, where waiting and overtime are
limited by probabilities. Chance constraint provides a natural
representation of Quality of Service parameters. It also promotes
fairness by accounting for individual surgeries and ORs. We consider a
set of surgeries with uncertain duration and a set of ORs with fixed
length of opening time, we decide (i) which ORs to open, (ii)
surgery-to-OR allocation and (iii) starting time of individual
surgeries, to minimize OR opening cost. We formulate chance constraints
to capture the on-time-start guarantees for individual surgeries and
the on-time-closure target of the entire OR sector. We provide two
alternative models where surgeries are scheduled in continuous time and
discrete time blocks respectively. Via sampling, we reformulate chance
constraints in each model as mixed-integer programming (MIP)
constraints based on finite realizations of uncertain surgery duration.
We also decompose the discrete-time MIP and develop cutting planes to
improve the computational efficiency. We test randomly generated
instances based on real data from a healthcare provider and develop
insights of how our approaches improve the quality of service and
fairness in stochastic OR planning.
36 Characterization
of Macrolesions Induced by Therapeutic High Intensity Myocardial
Contrast Echocardiography
Yiying I.
Zhu, Douglas L. Miller, Chunyan Dou and Oliver D. Kripfgans
cardiomyopathy (HCM) can lead to sudden death in young adults without
prior indication.  Echocardiography can be
used to diagnose HCM.  Current treatment is
highly invasive and inefficacious.  Contrast
echocardiography performed using therapeutic intensities is
hypothesized to induce sparse microlesions and treat the heart muscle

Contrast-enhanced pulsed ultrasound was employed to treat HCM with
electrocardiogram monitoring. Evans-Blue staining indicated necrosed
cardiomyocytes for microscope assessment.  To
reduce visual scoring ambiguity and provide a volume-oriented therapy
evaluation, a three-dimensional evaluation scheme was developed based
on brightfield (BF) and fluorescence (FL) tissue-sections (4096×4096,
16-bit, RGB). 

The treatment volume (macrolesion) was characterized with automatic
identification of microlesions on BF/FL images and a cost function to
penalize low SNR regions.  Currently 1:4096
downsampled data is processed in approximately 4 minutes, and requires
4 GB memory per CPU node. Thus, the original data could be processed in
2 minutes by 8000 nodes.

37 Molecular
Dynamics Simulations of Compressive Yielding of Cross-Linked Epoxies
Fasanella,Abhishek Kumar, Veera Sundararaghavan
dynamics simulations are performed to study compressive yielding
behavior of DGEBA/DDS cross-linked epoxy in a low temperature glassy
state. The molecular model was built using the dendrimer approach and a
representative epoxy structure was built by subjecting the structure to
several thermal annealing cycles. The simulations show that the
stresses drop sharply at the yield point, which was identified to occur
due to the activation of wedge disclinations of the epoxy chain. The
chemistry and geometry (critical segment length, angles, bond torsions)
involved in the molecular mechanism of compressive yielding have been
measured.  The results are analyzed in the
context of Argon theory, which is a linear elastic model of a wedge
disclination, and gives a critical kinking segment length that agrees
well with the molecular simulations. The yield stress versus
temperature predictions of Argon theory were directly compared with
molecular simulation results and show good correlation. Finally, the
use of Argon theory for extracting yield stresses at quasi-static
strain rates from high rate molecular simulations was investigated.
38 Monte-Carlo
Simulation Of Lamellar Grain Structure Arrangement using Discrete
Dislocation Plasticity
Ganesan, Dai Okumura
Aerospace Material
strengthening can be achieved by effectively obstructing the
dislocation motion. The mean grain size in polycrystalline materials,
d, plays a critical role in mechanical properties as the grain boundary
limits the free path of moving dislocations. In the present work,
Monte-Carlo simulation of polycrystal plasticity is performed using 2-D
discrete dislocation approach, to find the effect of grain size on λ,
the mean distance of dislocation activation across the grain boundary.
The approach is based on applying periodic homogenization to the
superposition method of Van der Giessen and Needleman, and on
decomposing displacements into macro and perturbed components. The
parallelized code is implemented to obtain λ for layer thickness of
1-10 μm for two different grain structure arrangements. The results
indicate that λ is very sensitive to grain structure arrangement and is
one order lower in lamellar arrangement when compared with other grain
structure arrangements and experimental results on pure Aluminum
39 Towards
Quantitative Dynamical Mean Field Theory
Li*Dominika Zgid
Chemistry In a
quantum mechanical description of the Kondo effect, the single spin of
the magnetic atom or molecule creates a loosely bound electronic
open-shell singlet with one of the electrons from the metallic surface.
A sharp resonance, the Kondo resonance, present near the Fermi level is
an experimental manifestation of this effect. The Kondo effect is
profoundly interesting because the arising magnetic behavior comes from
few unpaired electrons and, therefore, is of purely quantum
multi-determinantal nature and cannot be simulated successfully by
one-body approximations such as Hartree-Fock (HF) or Density Functional
Theory (DFT) with approximate functionals. In order to perform
quantitative calculations of the Kondo phenomena we use Dynamical Mean
Field Theory (DMFT) that can be understood as a QM/QM embedding method
and therefore allows us to correctly describe the coupling between the
magnetic atom d-electrons and the surface conduction band. We will
present newly developed updates to the DMFT method that allow us to
calculate the Kondo problem in an ab-initio manner without any
experimental parametrization. This approach will be tested on a series
of magnetic atoms and molecular systems deposited on a graphene in
order to calculate Kondo temperatures and photoelectron spectra.
40 How to make
dynamical mean-field theory applicable to large realistic systems?
Alexei A.
Kananenka, Dominika Zgid
Chemistry Transition
metal oxide perovskites are very challenging to describe due to the
presence of strong correlation among the orbitals. However, not all the
orbitals in these difficult systems are equally correlated. We aim to
design a computational method that could exploit some simplifying
features in the structure of electron correlation. To classify
different types of electron correlation we study the Hubbard model. We
propose a procedure for selecting the most important orbitals within
this model. In the basis of natural orbitals, the orbitals of the
two-dimensional Hubbard model are not occupied equally. This
observation is the basis of our method and we employ the modified
iterative perturbation theory (MPT) — an efficient and relatively
simple approach to find orbitals populations. Subsequently, for the
subset of the most correlated orbitals we use dynamical mean-filed
theory (DMFT) to provide a more accurate answer for the most important
orbitals in the problem. This procedure is general and in the future we
plan to apply it to realistic solids.
41 Robust
Algorithms for Limited Angle Imaging of Ultrasound Propagation and
Richards, Rungroj Jintamethasawat, Fong Ming Hooi, Paul Carson
Speed of
sound (SOS) and the acoustic attenuation coefficient are the simplest
two properties that can be extracted from breast tissue in the
mammographic geometry to improve cancer detection. The objective of
this project is to develop and employ robust algorithms for computing
SOS and attenuation images.  Limited angle
tomography of SOS imaging can be corrected for proper spatial
distribution and quantitative values by utilizing a covariance matrix
derived from a standard ultrasound image. Feature processing and image
reconstruction is computationally demanding, though the use of highly
parallel algorithms reduces the runtime to approximately 1.5 hours.  The employed shared memory model uses
approximately 1 GB of RAM per node for a total 300 MB of raw input data.  Both indicate relatively low demand on memory
resources but high demand on CPU. Calculated SOS in ultrasound phantoms
is within 1% compared to true SOS. Knowledge of SOS and attenuation
will improve characterization of benign and malignant tissues.