Organic semiconductor light-emitting devices have rapidly evolved, particularly through the development of OLED displays. This progress has continuously driven the search for higher-performance emissive materials with improved colour purity. In this context, organic lasers have attracted growing attention as next-generation display light sources capable of achieving the ultimate level of colour purity.
Beyond displays, organic lasers are also considered promising for future applications such as flexible photonics, optical wireless transmission (Li-Fi), on-chip optical communication, and low-cost coherent light sources. As a result, significant efforts have been devoted to discovering organic laser materials suitable for practical devices. One key advantage of organic materials is their compatibility with solution processing, which distinguishes them from conventional inorganic semiconductor laser systems.
However, discovering high-performance organic laser molecules is an extremely time-consuming process. Traditional research workflows typically involve designing molecules based on human intuition, estimating their properties through computation, and then synthesizing and purifying them one by one — a process that can take anywhere from several days to weeks for a single candidate. This does not even include the additional time required to fabricate and evaluate actual devices. Since researchers often need to repeat this trial-and-error cycle tens or even hundreds of times to identify promising candidates, the journey toward discovering high-performance materials can be long and challenging.
Moreover, such research heavily relies on skilled researchers. Experiments performed with careful attention to detail and accumulated know-how are essential for producing reliable, high-quality data. The challenge, however, is that the number of highly trained researchers is limited, while researchers themselves must simultaneously handle experiments, analysis, planning, writing, and many other responsibilities.
To address these efficiency challenges, self-driving labs (SDLs) have recently emerged as powerful platforms for accelerating materials discovery. Rather than being a single automated instrument, an SDL is better understood as an integrated research platform that combines simulation, robotic experimentation, data analysis, and in some cases even AI-driven candidate suggestion. By automating repetitive experimental tasks, researchers can focus more on molecular design, interpretation, and scientific strategy.
In this work, we adopted this SDL-driven approach to accelerate the discovery of organic laser molecules for next-generation organic semiconductor photonic devices.
Fluorene is one of the molecular fragments particularly well suited for this purpose. Because long alkyl chains can be introduced relatively easily, fluorene-based systems can achieve excellent solubility, which is highly advantageous for solution-process-based studies. In addition, fluorene possesses a structurally rigid framework and is known to exhibit relatively small excited-state reorganization energies, often resulting in favorable emissive properties.
Indeed, fluorene-based emitters such as F8BT and octafluorene derivatives have previously demonstrated excellent optical properties and laser activity. Nevertheless, the chemical space of fluorene-based organic laser molecules remains far from fully explored, making it an especially attractive target for SDL-driven high-throughput molecular discovery.
Motivated by these considerations, we focused on developing fluorene–acceptor–fluorene trimer structures. To connect the aryl fragments, Suzuki–Miyaura cross-coupling chemistry was employed using dihalide-based acceptor fragments and Bpin-functionalized fluorene building blocks. This design enabled one-step synthesis using a 2:1 coupling strategy.
Importantly, a wide variety of dihalide-based acceptor fragments are commercially accessible. This made it possible to rapidly combine and screen diverse acceptor structures, making the overall strategy highly compatible with SDL-driven accelerated molecular exploration.
The SDL platform developed by the Aspuru-Guzik group is particularly powerful for automating end-to-end in-solution workflows, including molecular synthesis, product isolation, and optical property measurements. In practice, the robotic system automatically performed repetitive experimental tasks including reagent dispensing, reaction execution, product handling, and optical characterization.
Our primary role was to design candidate molecules and organize experimental batches, while the SDL carried out the repetitive experimental procedures after batch execution.
First, a total of 252 experimentally accessible molecular candidates were designed. To focus on molecules with the potential for relatively long-wavelength emission above 500 nm — an area with comparatively limited examples — density functional theory (DFT) calculations were performed. Based on these calculations, 51 priority candidates were selected for experimental validation.
The SDL was then used to automatically synthesize and evaluate these 51 candidates. In practice, the system was capable of simultaneously synthesizing, isolating, and characterizing approximately 15 candidate materials per batch in a single day. Performing such experiments manually would likely have required continuous around-the-clock work or significantly extended experimental timelines.
Perhaps the most impressive aspect of the SDL workflow was that researchers could focus on designing new candidate structures or analyzing data while the repetitive experiments were being performed automatically. Watching experiments proceed autonomously overnight and checking the resulting data the following morning was a very different research experience compared to conventional laboratory workflows.
Among the synthesized molecules, two candidates exhibited particularly promising emission properties in the near-infrared (NIR) region above 700 nm. These molecules were based on diketopyrrolopyrrole (DPP) and benzoselenadiazole-based structures, respectively.
Achieving efficient emission in the NIR region is generally very challenging for organic molecules because non-radiative losses tend to increase significantly as emission wavelengths become longer. For this reason, organic laser materials operating beyond 700 nm remain relatively rare.
Interestingly, the structures identified in this study belong to molecular families that have been relatively underexplored in the organic laser field. To evaluate their laser potential, CBP-host-based thin films were fabricated and optically characterized. As a result, low amplified spontaneous emission (ASE) thresholds were successfully observed, demonstrating their promise as potential organic laser materials.
Personally, one of the most striking aspects of this work was experiencing the potential of SDLs firsthand. Experiments are essential for validating scientific ideas, but they are also often labor-intensive and time-consuming.
Although this study represents only one part of the broader challenge of developing practical organic laser devices, future stages such as device fabrication and long-term stability evaluation would require even more extensive repetitive experimentation.
At present, the range of experiments that SDLs can fully automate remains somewhat limited. However, if synthesis, characterization, device fabrication and evaluation, and AI-driven candidate suggestion can eventually be integrated into a single closed-loop workflow, researchers may be able to focus less on repetitive experimental execution and more on deciding what kinds of materials should be created in the first place.
Such a transition could fundamentally reshape not only laboratory automation, but the entire paradigm of materials discovery itself.