You can help push the Materials Science Frontier. Here's how:
Your survey responses directly feeds a Machine Learning model engineered for unparalleled speed in Electron Microscopy image reconstruction.
Electron Microscopy lets us take pictures of atomic structures, here's an example:
Except, this image itself is not from experiment, but a "perfect" simulated image of what we expect the atomic structure to look like. An actual experimental image looks like this:
Significantly lower quality, as you can see. I'm building an algorithm to fix these images, and importantly, to fix them fast enough for the algorithm to work within the microscope, reconstructing the image between the electrons hitting the detector, and the image being displayed on the monitor. Here's some preliminary results:
That's the enhancement of the really noisy experimental image shown above, and as you can see, it is significantly closer to the "perfect" simulated image.
An important knock-on effect of having this real-time reconstruction capability is being able to characterise materials that were previously too fragile for electron microscopy. This is because brightness correlates with specimen damage, but since we can afford to reduce the brightness knowing the algorithm will fix up the images for us, we will be able to finally image more fragile materials.
This includes novel two-dimensional (2D) materials, biological materials, materials for solar-cells and energy storage, and solid-state catalysts such as metal-organic frameworks (MOFs).
We will also be able to fine-tune image focus, specimen position, and other important parameters at the microscope because of the real-time feedback.