Shifting the paradigm of R&D

When scaling-up, nanoparticle synthesis becomes very costly and difficult due to the nature of the chemicals used to direct their formation, batch processing conditions and lack of adequate mass and heat transport. Conventional liquid-phase continuous techniques are also limited due to their low solids-handling capabilities. Top-down techniques like milling are also limited in their accuracy and purity.

The annular microreactor delivers enhanced control in bottom-up nanoparticle synthesis by harnessing a unique flow regime known as annular flow. In this configuration, a high-velocity gas stream forces liquids to form a thin film along the walls of a microchannel. This precisely engineered environment enables fine-tuning of nanoparticle properties.

Our patented design, is crucial for achieving consistent and uniform annular flow. This controlled environment ensures that reaction mixtures experience precisely regulated hydrodynamic forces and exceptionally thorough mixing. The result is the ability to tailor nanoparticle size, shape, crystallinity, and other key characteristics with a level of consistency and reproducibility that surpasses traditional methods. Additionally, the continuous reactor is resistant to the formation of clogs, making it one of the few microreactors that can withstand high solids-content mixtures.

Scaling up nanoparticle synthesis reactors presents a unique challenge beyond simply increasing volume. As reactor size grows, it becomes increasingly difficult to maintain the same precise flow behavior and mixing conditions achieved at a smaller scale. This variability can lead to inconsistent nanoparticle properties, rendering them unsuitable for the demanding requirements of advanced applications. Our solution tackles this problem by focusing on multiplying reactors instead of increasing their individual size.

A critical requirement for this approach is ensuring consistent flow distribution across many reactors, regardless of deviations in the reactors and downstream units. To address this challenge, we have developed a novel manifold to manage multiple reaction streams. This innovative solution accurately distributes the flow of reactants across multiple reactors simultaneously, ensuring that each reactor experiences identical and precisely controlled flow conditions.

While most systems rely on a complex network of sensors, “resistive” flow elements or feedback loops to achieve such consistency, our proprietary distributor manipulates flow in a simple yet robust manner that minimizes pressure drop.  The result is unparalleled consistency in nanoparticle size, shape, and other critical properties – enabling immediate scale-up without sacrificing the reproducibility achieved at the laboratory scale.

Our adaptive machine learning platform, offers a distinct advantage over established packages like scikit-learn or TensorFlow by combining a diverse toolkit of machine learning models with adaptive selection and tuning.  Rather than relying on a single approach, it intelligently assesses the complexity of a given nanoparticle synthesis problem and then chooses the optimization strategy.

This ranges from streamlined decision trees for simpler optimizations to powerful neural network ensembles for highly complex scenarios. An additional layer of machine learning then fine tunes the selected model parameters prior to training, ensuring peak performance without requiring specialist data science expertise.

Its multi-objective capabilities enable simultaneous optimization of factors like yield, cost, and morphology.  Moreover, by intelligently selecting the optimal strategy for each objective, the algorithm delivers tailored solutions for specific nanomaterial goals.

Complex nanoparticle systems are often poorly understood and demand a sophisticated optimization approach. With only probabilistic, Bayesian models at its core, the algorithms strategically prioritize unexplored regions of parameter space, leading to faster, more efficient optimization.

Our solution was tested against a popular multi-objective ML package. In the synthesis of a complex multi-step, multi-objective chemical system, our solution exhibited 60% higher yields, at a lower cost, when given the same number of experiments. With its focus on flexibility, efficiency, and handling multiple objectives, our approach offers a practical approach to accelerating research and minimizing costs, even when compared to other sophisticated ML solutions.

All of our software development is performed using our bespoke automation framework. Existing automation solutions generally fall into two categories. Conventional solutions, are limited to simplistic automation routines without AI, and are often tied to proprietary hardware.

Our framework was developed to address these limitations. Built in python, it provides a modular, universal approach to integrating devices, concurrent tasks, AI models, user interfaces and data pipelines. This allows us to quickly create simple yet powerful workflows incorporating automated devices and machine learning methods – known as autonomous or “self-driving” systems.

We’ve built a library of high quality drivers for diverse hardware, from pumps and valves to robotic arms, with an infrastructure that allows for flexible expansion to incorporate additional devices in the future.  ITtalso enables the orchestration of complex workflows involving multiple devices, data and inputs simultaneously.  Additionally, robust data pipelines and structures ensure organized integration of data across devices, machine learning systems, and external repositories from which users can import and export data.