Home

Alexander Couturier

I am a Ph.D. physicist with a unique background spanning:

Laboratory research (UV fluorescence spectroscopy, ellipsometry, QCM-D, chemical wet lab solution prep, nanoscale sample prep, polymers, material science, AFM, SEM, optics, nuclear engineering, high power pulsed lasers, plasma)
Scripting/Programming (Python, MATLAB, C, C++, Java, Arduino, JavaScript, UNIX, Git, fundamentals of AI & ML)
Graphics (Adobe Suite, CAD, Autodesk Maya, particle simulations)

Get a closer look at some of my past projects below...

QCM Transfer-Matrix Model

I developed a new technique for computationally handling large-scale matrix operations to analyze quartz crystal microbalance data

This enabled the discovery of a localized nanoscale variation in complex shear modulus that could not previously be found

The work had major implications for the science of how material properties couple between two adjacent media

Click here to read the paper on this

Click here for GitHub repository

Spectroscopy Equipment Diagnostics and Optimization

On three separate occasions, one of which was during my first few months in the lab, I diagnosed and fixed existing faults that were causing subtle shifts in the collected data

Other lab members with more experience at the time did not catch these faults

Furthermore, I designed thorough tests and wrote scripts to aid in optimizing the collection and analysis of the data

This yielded a 40% reduction in variability of results

Interactive Data Fitting Tool

Fitting the glass transition temperature (Tg) from fluorescence spectroscopy data comes with unique challenges that needed solving

It is done by fitting two lines and finding the intersection, which is very slow to do manually over and over in trying different fitting ranges
It is best to have your choice of fitting ranges be informed by fit line error, but you don't get that ahead of time with conventional approaches to fitting

The solution to this dilemma was a custom tool that incorporates a JavaScript callback in Python to provide real-time feedback on your choice of fitting ranges

Click here for GitHub repository

Discovery of a New Confinement Effect

I pushed the limits of our nanoscale sample preparation protocol, leading to engineering features at the level of < 100 atoms

This, combined with the above spectroscopy improvements, enabled important revelations about polymer confinement, reconciling decades of seemingly contradictory literature

Click here to read the paper on this

Novel Phenomena in Machine Learning

I led a team of machine learning researchers on a project investigating novel double descent test error behavior using different techniques and parameterizations

We employed polynomial regression, neural nets, and support vector machines

(double descent only ultimately observed with regression)

Regression

Singular value decomposition with parameter spaces constructed from principal component analysis as well as Legendre polynomial orthogonalization

Neural Nets

Fully connected neural network with a rectifier-activated input layer and a fixed hidden layer that had a variable number of neurons of sigmoidal activation

SVM

Used polynomial kernels of varying degree, and compared results with that of kernel ridge regression

Chemical Wet Lab, Ellipsometry

I have experience preparing hundreds of chemical solutions from a range of solvents and polymers

Experienced in safe handling of corrosive, flammable, and explosive chemicals

I have also prepared countless nanoscale films supported on silicon via spin-coating (left)

Primary sample geometry characterization was performed using ellipsometry (right)

Additional sample prep protocols have included:

Etching using piranha solution (sulfuric acid + hydrogen peroxide)
Polymer multilayer stacking using different hydrophobic/hydrophilic water transfer techniques

Image Classification

I designed abstract classification features for perceptron image recognition of handwritten numbers

These features were simple yet effective at enabling a high success rate in recognizing the correct digits

MATLAB 3D Heatmap

Animated 3D heatmap done in MATLAB for displaying a scalar field changing along a surface

Graphical Interactive Learning App

During my education in computer graphics technology, I led a team in collaboration with the local historical association to build an interactive tool they could have visitors to their exhibit use

Recreation of Fort Ouiatenon from the 1700s

Built in Unreal Engine, with assets made primarily in Autodesk Maya

The idea was that visitors could see what this historical location looked like hundreds of years ago, and they could click individual items to get a display of detailed information

Nanoscale Imaging

Imaging graphene flakes under a microscope (left) was a big part of one of my projects

The goal was identification of monolayers to use for studying the quantum Hall effect

AFM imaging of ultrathin gold films (right) I made via sputtering deposition

I was aiming to minimize surface roughness for studying two-photon photoluminescence

Product Design

CAD software (CATIA)

Java Statistical Simulation GUI

Click here for GitHub repository