Student Posters Repository

My research is to assess the safety of rocket on liquid fuel tank destruction. In case of failure of rocket launch, flight termination system is activated to prevent the resulting ground damage due to the induced hazards such as explosions. In a liquid rocket termination operation, fuel is ejected to the mainstream. The self-ignition of the ejected fuel can occur due to the interaction with high speed mainstream. In this research, to clarify the criteria of self-ignition and flame-holding, the investigation of mixing state is performed by using two-dimensional and three-dimensional numerical simulation. As a result of the investigations under wide-range flight conditions, it is founded that the mixing mechanisms are affected by the momentum flux ratio and the pressure ratio governing the size and shape of jet plume. Furthermore, it is confirmed that the specific vortex at high temperature and low velocity has adequate fuel, it can induce self-ignition and flame-holding.

Understanding human impact on climate is the foremost challenge of the 21st century. In particular, significant work remains to correctly model the multi-scale nature of clouds. In particular, the evolution of cloud systems is a combination of the small-scale "microphysics" of liquid droplets and ice crystals and the large-scale atmospheric flows that supply moisture. Several examples of techniques to use cloud simulation and satellite data are presented. First the adjoint model of ice nucleation codes is used for global-scale attribution analysis, in which we identify the input variables that control temporal variability in cloud ice crystal number concentrations. We then zoom in, tracking a single control volume of atmospheric air to understand how thermodynamic conditions affect the evolution of these ice crystals numbers. Under the right dynamic conditions, these individual clouds may also organize to form mesoscale systems, and preliminary results of the effect of surface warming on this organization are shown. Future work will focus on using information entropy metrics to more rigorously quantify the degree of this clod aggregation.

Carbon dioxide, Ozone, Water vapor etc are common examples of atmospheric trace gases. Accurate representation of such tracers in climate models is essential to obtain accurate climate predictions. While most models, using different numerics and grids, get the simulated global mass transport correct, there is still a huge spread in tracer transport in such state-of-the-art models. In this study, we create a mathematical framework, to understand and resolve these differences in tracer transport across models. We consider modern dynamical cores(used for climate and weather prediction), developed at Princeton-GFDL and NCAR, that use a wide range of numerics(Pseudospectral, Finite Volume and Spectral Element) and use an idealized tracer called age-of-air to compute the mean residence times of tracers in the stratosphere and create a benchmark study for climate dynamical cores.

Many real-world examples of fluid-structure interaction, such as the motion of red blood cells through the narrow slits of the spleen, involve the near-contact of elastic structures separated by thin layers of fluid. The separation of length scales between these lubrication layers and the elastic structures themselves poses significant computational challenges. To efficiently handle this multiscale problem, we introduce an immersed boundary method that uses elements of lubrication theory as a subgrid model to resolve thin fluid layers between immersed boundaries. We apply this method to flows of increasing complexity to show its increased accuracy compared to the classical immersed boundary method. We present preliminary simulations of cell suspensions, in which near-contact occurs through cell-wall, cell-cell, and intracellular interactions, to highlight the importance of accurately resolving these thin fluid layers.

The Atlantic hurricane season runs from June 1st to November causing massive destruction and loss of life. In 2017, 17 named storms hit the Atlantic causing destruction worth an estimated $316 million and at least 464 fatalities. Meteorologists, by studying previous weather data, predict the expected number of hurricanes in the season. These predictions help authorities prepare for disasters and over the years, better predictions have minimized loss of life and property. However, these predictions rely on human expertise and are often extremely complex due to the thousands of parameters involved and the chaotic nature of weather.

The aim of this study is to introduce Machine Learning to hurricane prediction. Recent scientific advancements have seen geostationary satellites capable of collecting tens of Terabytes of daily data of the weather. On the other side, machine learning models propose efficient techniques to analyze such large data and extract meaningful information. With this large amount of data and the power of high performance computing, machine learning could be an alternative tool for climate study.

Machine learning has the ability to understand complex models and relationships in data. Recent developments in deep learning models such as Deep Neural Networks (DNN) have led to significant achievements in accuracy. We introduce machine learning model to hurricane prediction to explore the complex relationship between multiple factors such as sea surface temperature, sea level pressure, sea ice cover and wind patterns. We aim to apply a deep learning model to understand the effect of these parameters in the hurricane season and the number of hurricanes. Such insights could significantly improve disaster preparedness and give authorities a better picture of what to expect in the hurricane season.

Using historical data, we train a DNN to classify the hurricane season based on the number of hurricanes likely to occur. Our initial experiments are aimed at finding the most significant factors in storm formation. Preliminary results show that Sea surface temperature has the highest impact on the prediction of the number of storms. Furthermore, given the average sea surface temperatures in a month, our DNN model is able to predict the number of hurricanes with about 60% of accuracy.

Our future goal is to develop a complete end to end workflow to continuously learn weather patterns that affect the hurricane season and accordingly, to make predictions. We plan to implement distributed learning using PyCOMPSs (a programming model and runtime which aims to ease the development of parallel applications for distributed infrastructures, such as Clusters and Clouds) to reduce or eliminate the need to expensive computational infrastructure in climate science.

Rigidity percolation occurs when mechanical stability emerges in disordered networks as more components or constraints are introduced. Classical theories of rigidity percolation elucidated critical phenomena at the rigidity percolation transition in systems where components are uncorrelated. Many experimental systems, such as colloidal gels, involve components that exhibit interactions which induce positional correlation. In this paper, we discuss the effect of correlation at the rigidity percolation transition and discuss its implications for colloidal gels. We find, through numerical simulations of site-diluted triangular lattices, that short-range positional correlation shifts the rigidity percolation transition to lower volume fraction, while keeping the same critical exponents, consistent with the scenario that correlation acts as an irrelevant perturbation at rigidity percolation. We further explore the emergence of rigidity in colloidal gels through molecular dynamics simulations and structural rigidity analysis. In particular, we examine the relation between the emergence of rigidity and the gelation transition at different temperatures, aiming at understanding the origin of structural rigidity in colloidal gels, as solid materials at extra-low volume fractions.

Over the past several decades, computational fluid dynamics has advanced rapidly, and the range of available numerical algorithms and computationally feasible physical problems has expanded. Modern numerical solvers provide a compelling opportunity to probe for as-yet undiscovered effects that emerge with longer integrations and higher numerical precision. In this study, we first derive a range of linear modal instabilities in self-gravitating disks. Improved resolution permits identification of slowly growing instabilities, which may have a close connection to structures observed by ALMA. We use our linear solutions to the hydrodynamic governing equations to assess the fidelity of meshless and conventional grid-based hydrodynamic solvers. We then study the weakly nonlinear long-term development of endogenously developed spiral modes. By comparing modern simulations with prior results, we hope to provide a stronger understanding of the impact of fluid mechanics upon the evolution of protostellar disks.

sorry, had no to write an abstract :P

Description of electron correlation is crucial to chemical accuracy in quantum chemical calculations. However, the interaction of electrons in a system is an insoluble many-body problem. Constructing approximations to describe electron correlation is a challenging task, but there has been success in the quantum chemistry community using wavefunction methods and density functional theory. However, there are still challenges to be overcome in areas such as theoretical solid state chemistry, which requires description of large systems and the use of finite temperature. Large system sizes and finite temperature can be difficult to treat solely with the currently available methods. Therefore, a new class of methods, based on the temperature-dependent Green's function, is implemented and explored. 

This work is toward investigating the use of temperature-dependent Green's functions for ab-initio quantum chemistry. That is, we are treating a quantum chemical Hamiltonian with realistic electron interactions. While this formalism has been applied to model systems in the condensed matter community, its has been used much less by the quantum chemistry community. Therefore, the numerical behavior and accuracy of Green's function methods for quantum chemical calculations is relatively unknown.

This work investigates the ability of the temperature-dependent Green's function, which is an ensemble formalism, to give access to temperature dependent thermodynamic quantities such as free energy and entropy when calculated in a second-order and perturbative manner (GF2). We find that this method is able to give good accuracy for lower temperatures and excellent accuracy for higher temperatures for a molecular case and is able to qualitatively describe a simple model of a solid. The results of this work are presented in chapter 3. 

Although Green's functions have a clear connection with spectra at zero temperature, it is not straight forward to obtain spectra from the finite temperature Green's function, which is calculated on the imaginary frequency axis. Therefore, we must investigate methods to obtain spectral quantities in a consistent and reliable manner from the imaginary axis. Chapter 4 investigates several methods to do so and we compare our results with experimental and highly accurate benchmark data. We find that it is possible to obtain spectral quantities that can differ by several electron volts, even if the same level of theory is used to obtain the Green's function. This reiterates that finding a spectrum from the imaginary axis is nontrivial and that one must exercise caution when comparing spectral quantities that were calculated using different techniques, even if they are treated with the same theoretical accuracy.

Accessing larger systems with the Green's function requires the use of quantum embedding. Quantum embedding describes nontrivial electron interaction between a highly accurate ``active space'' or ``impurity'' and a larger, lower level ``environment''. It is challenging to construct an impurity solver that is reliable at low temperatures. In chapter 5 of this work, we implement and test a temperature dependent configuration interaction impurity solver for quantum embedding. This solver can be used in quantum embedding schemes such as dynamical mean field theory and self-energy embedding theory for larger systems. 

Overall, this work has made progress toward using Green's functions for ab-initio quantum chemistry at finite temperature. Groundwork has been laid for using this formalism to calculate thermodynamics and spectra using a Green's function with realistic electron interactions and to explore quantum embedding using an impurity solver at low temperatures. 

Image processing of faces and topic modeling of text are both widely used forms of processing employed in machine learning applications. Facial features can be used to identify faces that look similar or in an ideal case, a specific person, depending on the quality of the image. Topic modeling of text articles is employed across a variety of applications to improve content search from news or research articles, and is also used in a variety of natural language processing applications such as chat bots. The purpose of my research is to identify and validate candidate methods for linking textual articles and images they contain for assessing previously unseen images and/or articles. This requires that I transform both target data types (images and articles) into their respective n-dimensional spaces. I will use both convolutional neural nets (CNNs) and self-organizing maps (SOM) to help me encode the different instances (image, article) and store them for indexing. Data size and computational intensity require HPC for this to be successful. My research also explores identification of a common n-dimensional functional space where vectorized images and text can exist together, enabling classification of new articles and images based on the trained article and image pairs.