IMID 2026
Date & Time - August 18th-21st, 2026
Location - Busan, South Korea
Schrödinger is excited to be participating in the 26th International Meeting on Information Display conference taking place on August 18th – 21st in Busan, South Korea. Join us for a presentation by Mathew D. Halls, Senior Vice President, at Schrödinger, titled “Discovery of Novel Display Materials and Devices Using Machine-Learning Accelerated Simulations and Generative AI.” Stop by our booth 27&28 to speak with Schrödinger scientists.
AUG 21 | 12:05 PM 
Room F (313) Discovery of Novel Display Materials and Devices Using Machine-Learning Accelerated Simulations and Generative AI
Speaker:
Mathew D. Halls, Senior Vice President, Schrödinger
Abstract:
The rapid evolution of organic light-emitting diode (OLED) technology requires the simultaneous optimization of molecular emitters, host materials, and complex device architectures. Traditional Edisonian trial-and-error approaches are increasingly insufficient to navigate the vast chemical and structural space required for next-generation displays. This work highlights three major advancements in the concept of an “atom-to-device” design framework for OLED applications, dramatically accelerating the pace of innovation in display technology (Fig. 1).
1. Generative AI for Molecular Design: The REINVENT framework is an undirected generative model trained to explore novel chemical design spaces. By applying multi-objective reinforcement learning within a target property space, one can generate synthetically viable materials candidates with optimized optoelectronic properties for display and lighting applications.
2. Machine-Learning Force Field (MLFF) with MPNICE Framework: To overcome the “static snapshot” limitations of computationally expensive methods such as DFT, we have implemented highly efficient machine-learning force fields using the MPNICE framework. This new class of force fields enables dynamical simulations of molecular systems at the time and length scales previously inaccessible to conventional electronic-structure-based methodologies. We demonstrate that the MPNICE framework also provides a path to navigate high-dimensional potential energy surfaces for optimization problems where ab initio techniques become impractical for larger scale OLED applications.
3. Machine-Learning Device Algorithm for Tandem OLED Devices: We extend Schrödinger-exclusive machine-learning framework for OLED device design to tandem OLEDs. The new workflow solution is specifically designed to predict the performance of tandem devices with respect to key engineering design factors such as layer thickness and material properties, facilitating the optimization of complex stacked architectures from the atomistic scale up.
These new capabilities for optoelectronic materials development and device optimization provide unprecedented tools for the accelerated development of innovative display technologies. The integration of generative AI technology for chemistry with physically grounded MLFFs and device-level predictions establishes a new paradigm for the “digital-first” development of high-efficiency, long-lifetime display technologies.
