Since the 2004 discovery of graphene, numerous 2D quantum materials have been synthesized, with vast potential to revolutionize condensed matter physics and electronic technology by providing easy access to quantum phenomena. As the ability to make these systems has grown, the primary challenge has begun to shift to understanding the complex phenomena that occur in 2D materials. The QMO lab has focused on innovating new ways of seeing into these systems, combining techniques from several disciplines, including optics, automation, and image analysis to develop a novel technique, called Multi-Parameter Dynamic Photoresponse Microscopy (MPDPM) that can reveal the full range of optoelectronic behavior exhibited by a quantum material using data intensive imaging.

MPDPM differs from other experimental methods by treating the complex photoresponse of 2D materials as a data problem. The internal complexity and the indirect nature of quantum mechanics pose significant challenges in understanding 2D systems. Normally this is treated as an experimental problem, cleverly setting up the experiment to cut through the complexity and capture the quantum phenomena in as simple a data set as possible. But as more complex 2D systems are developed, this is harder to do and has more drawbacks. Alternatively, MPDPM embraces the complexity, seeking to systematically image the photoresponse of a sample over a wide range of experimental variables. Then, once a large comprehensive data set has been gathered, image analysis, statistical analysis, and various other data mining techniques, often borrowed from high energy particle physics or astronomy, are employed to extract results from the data set. This transforms an issue of experimental complexity to an issue of data analysis.

Figure 1 shows a simplified schematic of how the data analysis process works by reducing a large set of images to a single visualization. To explore the behavior of an example graphene based heterostructure we efficiently acquire a large set of 625 photocurrent images that fully cover the response of the sample to excitation by laser light. Then we convert subsets of the images into processed images that represent the behavior due to an experimental variable, in this case laser intensity. Then we identify and isolate a physically interesting feature from the processed images, in this case the “edge” around a bright area and put it into a single visualization, in this case by plotting it on a microscope image of the sample. This final visualization is much more scientifically useful that 625 images and, if the identified feature is chosen well, encapsulates the behavior of the sample. Exactly what sort of image processing, feature identification or visualization is used will depend highly on the experiment, but the key to this technique is to efficiently acquire a large set of images that capture the behavior of a sample over multiple variables then use image processing and analysis to condense the data set into useful visualizations.

The QMO lab first developed this technique to reveal the formation of an electron hole liquid in 2D MoTe₂. Since then, we have learned to utilize this technique in almost all of our experiments, including studies of Heterojunctions, Thermal Energy in Graphene and Thermospintronic imaging. The technique is so useful because it allows us to get the most out of the limited number samples that we can fabricate. However, the core concepts are not specific to nanoscale physics and have many potential applications in other areas. For example, in biophysical systems complexity is inherent and impedes our understanding of important biological processes such as photosynthesis. The QMO lab is developing data intensive measurements similar to MPDPM for biological systems that may help us understand the growth of photosynthetic bacteria, the thermal vision of pit vipers or the flow of energy in complex biomolecules. More generally, any complex phenomena where data can be rapidly acquired may benefit from the data intensive methodology similar to MPDPM.