From Automated Light Scattering to Autonomous Material Design

From Automated Light Scattering to Autonomous Material Design

Share this post

Choose a social network to share with, or copy the shortened URL to share elsewhere

This is a representation of how your post may appear on social media. The actual post will vary between social networks

Most functional soft materials are complex fluids where one phase is evenly dispersed within another, like cream or foam. These materials often appear stable on a large scale, but at the microscopic level, they exhibit constant, random, and spontaneous movement (like Brownian motion). Under certain conditions, these spontaneous dynamics drive the phases within complex fluids to separate. During phase separation, microscopic objects organize into larger intermediate structures that dramatically impact the material's overall properties – a good example is the way soy milk transforms into tofu. Understanding the conditions that trigger phase separation, along with the pathways of spontaneous dynamics in complex fluids, is crucial for designing functional soft materials with tailored characteristics.

Directly measuring spontaneous microscopic dynamics is challenging. Consider a vial of colloidal nanoparticles suspended in water: billions of nanoparticles can change position due to Brownian motion in under a millisecond. It's impossible to track the position of every single particle at every moment. Instead, we use light scattering to study the statistics of the dynamic behavior of these objects across various spatial and temporal scales, determined by the wavelength of light and the detector's time resolution. This statistical information reveals thermodynamic properties and offers physical insights when comparing dynamics across different experimental conditions and sample compositions.

Two common light scattering techniques are Small-Angle X-ray Scattering (SAXS) and Photon Correlation Spectroscopy (PCS, also known as Dynamic Light Scattering or DLS). SAXS uses sub-nanometer wavelength X-rays to measure the time-averaged spatial arrangement of mixed-phase structures with nanometer-scale sensitivity. PCS, with visible laser light (> 300 nm wavelength), measures temporal fluctuations of these spatial arrangements with tens-of-nanosecond resolution using a single photon detector.

Small-angle X-ray Photon Correlation Spectroscopy (SA-XPCS) combines the strengths of SAXS and PCS. It probes the spontaneous dynamics of mixed phases with both nanometer-scale spatial sensitivity and sub-microsecond time resolution, thanks to the use of coherent X-rays at synchrotrons/free-electron-lasers and the development of million-pixel, single-photon-counting X-ray detectors capable of megahertz frame rates.

Typically, the throughput of XPCS is limited by available coherent flux since only about 1-10% of the X-ray beam produced by 3rd generation synchrotrons is coherent. However, new 4th generation synchrotron sources like MAX-IV in Sweden, the ESRF-EBS in France, SIRIUS in Brazil, and the upcoming APS-U in the United States will boost the usable coherent beam fraction to nearly 100%. This potentially offers a 10,000-fold increase in XPCS measurement throughput, taking sample turnaround times from days or hours down to minutes or even seconds. Self-driving lab capabilities are therefore essential to utilizing the full capability of these billion-dollar scientific investments.

The development of a self-driving lab progresses through three stages:

  • Automation: Robots precisely execute well-defined experimental protocols, ensuring repeatability.
  • Autonomy: AI analyzes results from the previous N experiments, determines the goals of the N+1 experiment, and directs the automation robots to perform it.
  • Abstraction: Scientists describe the material's desired properties verbally. The AI translates these into measurable physical quantities and designs the automation protocols needed to carry out the experiment.

In the recent paper “Robotic pendant drop: containerless liquid for μs-resolved, AI-executable XPCS” published in Light: Science and Applications, Ozgulbas et al. demonstrate using containerless liquid (like a pendant drop) as a sample holder in SA-XPCS measurements. Coherent X-ray scattering intensity is captured by an ultrafast pixelated photon-counting detector (Rigaku XSPA-500k) synchronized with the synchrotron clock frequency at the Advanced Photon Source (APS) to achieve 3.7 μs time resolution. An electronic pipette mounted on a 6-axis robotic arm (UR3e) generates the pendant drop and handles automated liquid sample loading and disposal. A tool changer on the robotic arm allows switching between various end effectors (pipette, gripper, etc.), offering the flexibility to perform diverse scientific tasks using a single robotic arm.

Figure 1. Robotic pendant drop at APS. (Left) “Digital Twin” in Nvidia Isaac Sim. (Right) Setup in the adjacent chemistry lab of Beamline 8-ID. The inset shows the optical image of the drop captured using an inline optical camera. Details for all labelled parts can be found in the Method section of Ozgulbas, Light Sci Appl 12, 196 (2023).

This robotic pendant drop represents a step towards a fully-automated light scattering experiment with unattended liquid sample preparation. Beamline 8-ID-I has also developed automatic, high-throughput XPCS analysis (Zhang et al., J. Synchrotron Rad. [2021]. 28, 259) and AI-driven interpretation of XPCS results (Horwath et al., arXiv:2212.03984 [cond-mat.mtrl-sci]). Combining these technologies paves the way for autonomous materials discovery. We have also demonstrated that the robots can be equipped with depth-sensing cameras (Video 1), which allows the robots to identify and interact with their surrounding environments. This provides the possibility of creating a “Digital Twin” in the virtual world that can update according to the physical reality. By generating discrete robot motions using modularized robotic programs such as the Workflow Execution Interface (WEI), and combining with AI tools such as linguistic model and pattern recognition, researchers will be able to leverage the abundant automation and computing resources at U.S. DOE National Laboratories to achieve abstractive material design goals without having to go through the lengthy technical cumbersomeness of sifting data or programming robots. This “self-driving lab” concept will significantly improve the bandwidth of scientific facility users, and the protocols that are developed at scientific facilities will also benefit smaller-scale laboratories in universities and industries.


Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Follow the Topic

Control, Robotics, Automation
Technology and Engineering > Electrical and Electronic Engineering > Control, Robotics, Automation
Materials Characterization Technique
Physical Sciences > Materials Science > Materials Characterization Technique
Synchrotron Techniques
Physical Sciences > Physics and Astronomy > Atomic, Molecular and Chemical Physics > Synchrotron Techniques