NCSI: Improved Force Microscopy Image Reconstruction through Numerical Simulations and Modeling

Status	Completed
Mentor Name	Ryan Dwyer
Mentor's XSEDE Affiliation	Startup Allocation
Mentor Has Been in XSEDE Community	1-2 years
Project Title	Improved Force Microscopy Image Reconstruction through Numerical Simulations and Modeling
Summary	In force microscopy, magnetic or electrostatic forces between a nanometer-scale tip and a sample surface can produce nanometer-resolution images; however, linking the raw experimental data to specific sample properties can be challenging. Using our recently developed Lagrangian impedance model, we will perform numerical simulations for a broad range of possible tips and samples and apply several image reconstruction methods to the simulated data. The student will determine which image reconstruction method best captures relevant sample electrical or magnetic properties and share these results with other force microscopists via open-source Python code and a web interface for those without programming experience.
Job Description	The student will start by running numerical simulations of force microscopy experiments using an existing Python codebase. To efficiently perform simulations covering a broad range of different tip and sample properties, the student will write scripts to perform these analyses using the Ohio Supercomputer Center. To efficiently use many cores, the Ray Python package will be used for parallelism. This part of the project will generate a large quantity of simulated experimental data. The student will write scripts to analyze and visualize the results of these simulations. From this analysis, the student will draw conclusions about which experimental conditions best distinguish different sample properties. Finally, the student, with the help of the faculty mentor, will apply existing image reconstruction algorithms to the simulated data. The student will analyze these image reconstruction methods to see which methods most faithfully reproduce sample properties under the broad range of experimental conditions they have simulated. The student will make these results accessible to the broader scientific community by posting one or more of these image reconstruction algorithms publicly on the web using the Streamlit Python package and XSEDE Jetstream resources, which will enable other researchers to apply this analysis to their own images with no coding on their part. The student will summarize their results in a technical report and an oral or poster presentation. The technical report and presentation will be used to create an abstract suitable for submission to a scientific conference, such as an American Chemical Society regional meeting.
Computational Resources	We have access to low-cost computational resources through the Ohio Supercomputer Center; those will be the primary computing resources used. When appropriate, XSEDE resources will be used for training. The web interface to make these force microscopy analysis tools available to the broader community will use XSEDE Jetstream. If XSEDE Jetstream resources are unavailable, we will substitute a more limited, free platform such as Heroku.
Contribution to Community	A new student researcher will learn Python programming, data visualization, and high-performance computing. These skills will allow the student to contribute to XSEDE and its mission. The student researcher will also develop skills in scientific writing, communication, and presentation. These skills are also essential to ensure that high-performance computing reaches the widest possible audience. Scientifically, force microscopy is used by a large community of scientists who are often excluded from the latest advances in theory and technique because they lack the time and/or expertise to learn and use these new approaches. By sharing our key results through a web interface that requires no coding, this project will make recent advances in force microscopy modeling and image reconstruction more accessible to other researchers and users in the field.
Position Type	Learner
Training Plan	The student will learn the necessary Python programming skills using Software Carpentry, Learn Python the Hard Way, and other resources provided by the mentor to help the student learn the existing Python codebase. Along with these programming skills, the student will learn the basics of scanned probe microscopy by reading some of the relevant scientific literature and attending a low-cost virtual scientific conference (the ACS Great Lakes Regional Meeting). The student and mentor will have regular will have regular virtual meetings to discuss results, questions, next steps, and ensure the student has clear goals and tasks to perform. With a basic background in Python and scanned probe microscopy, the student will gain familiarity with the existing codebase by reproducing existing analyses on their own computer with guidance from the faculty mentor. This training will help the student develop a feel for performing “computational control experiments” to verify their results as they go. From there, we will move on to using the Ohio Supercomputer Center to perform these analyses in parallel more efficiently. To introduce the student to high-performance computing, we will make use of the outstanding training resources provided by the Ohio Supercomputer Center (and XSEDE where appropriate). Software Carpentry resources will also be used to help the student become more comfortable with the command line and shell scripting. These training resources will help the student learn how to use shell scripts and command line tools to submit batch jobs to the supercomputer, analyze data, and automate repetitive steps of the computational workflow.
Student Prerequisites/Conditions/Qualifications
Duration	Summer
Start Date	06/01/2021
End Date	08/06/2021