Tell me, wouldn't you like some magic in your life? What if I told you, it can be created from devices so small that we can fit 10000 of them across the diameter of a single human hair. These nanodevices are composed of tiny nanomaterials and have the potential for fascinating applications. 

Imagine a chip that mimics the retina of an eye but, unlike a human eye, can help us identify light from distant galaxies, take images of neuronal activity in the brain, and provide us new forms of secure communication. At Berkeley Lab, we are building such a chip, which will host arrays of nanodevices capable of detecting light with extreme precision; at single photon level. And to that, we'll use engineered nanomaterials called carbon nanotubes. 

But as you can imagine, constructing such devices requires high precision, substantial resources, and several design cycles. What if we had a 'crystal ball' that predicts the performance of our device, like a plot of current versus voltage, without us spending time and effort constructing every single device? That's where computational tools come in, and my role is to develop these tools to help scientists design such chips. 

But the problem is, even if you make your tools very accurate, they may predict results that are far from experiments—like my old simulation. 

So, what's missing? It's too idealized. The key here is to ask the crystal ball the right kind of questions. And the question I like to ask is whether my simulation setup really agrees with the realistic experimental conditions. For instance, there could be external/foreign charges around nanotubes introduced during experiments, unknowingly, that could affect their performance. Such non-ideal effects have so far been completely ignored in computational studies, but my tools show that they may have a significant impact on performance -- like my new simulation; with these effects added, it’s much closer to the experiment.

But the what's the catch, you may ask? The I try to get to realistic conditions, higher the computational costs. That’s why I strive for faster computations and utilize Perlmutter supercomputer at NERSC, here at the lab, which allows me to distribute my computations across thousands of mini-computers. In essence, I am harnessing the power of modern chips to help design future chips. Wait, that’s very meta, isn’t it?

What's next? I would like to extend the application of my computational crystal ball to model far more complex nanodevices, so that we can all tap into their true potential -- to make our world feel – magical.