A general protocol for phosphorescent platinum(II) complexes: generation, high throughput virtual screening and highly accurate predictions

A research team from the University of Hong Kong has developed a general computational protocol for phosphorescent platinum(II) complexes via high throughput virtual screening and Δ-learning approaches. By generating a large quantity of candidates, rapid HTVS and accurate predictions of vital photophysical properties, this innovative approach has resulted in 12 promising candidates with exceptional performance. This study paves the way for a faster, more reliable and more prosperous future in OLED technologies.

Cyclometalated Pt(II) complexes are widely used triplet emitters due to their color-tunable emissions. To make them viable for practical applications as OLED emitters, it is essential to develop Pt(II) complexes with high radiative decay rate constants and photoluminescence quantum yields. Despite the advantages of first-principles calculations, the results derived from density functional theory (DFT) or time-dependent DFT (TDDFT) calculations on phosphorescent platinum emitters are not accurate compared with the experimental ones. Furthermore, the computational costs are usually not affordable for a large library of metal complexes when exploring vast chemical space. Despite the investments in high-throughput virtual screening (HTVS),  the Pt-based PhOLED field still faces major challenges. One key issue is the limited availability of potential structures in public datasets, which restricts the discovery of novel compounds. Additionally, the diversity of coordination bonds creates complex representations, making it harder to model these systems accurately. Another vital issue lies in integrating HTVS with machine learning (ML) models. The intricate nature of coordination chemistry further complicates data-driven predictions, requiring more sophisticated algorithms and better-curated datasets.

The Solution: The researchers reported a new protocol of platinum complex generation, HTVS, and highly accurate prediction to screen promising phosphorescent Pt(II) emitters. Based on the core structures and the functional groups, more than 3600 synthesis-friendly complexes were generated. A feature strategy concerning core structures, molecular properties and testing medium is established to represent each complex correctly and validly. To speed up the screen of candidate molecules, different ML algorithms were used and compared based on a dataset of 198 Pt-based emitters. Recently reported Pt-complexes were also introduced as external samples to evaluate the capability of the HTVS-ML models, which indicates the generality of the optimal models. Under the favor of a three-tier well-designed screening rule and accurate predictions approach, 12 most promising complexes were selected and recommended for further synthesis. This work presents the first ML-based general protocol for generating, high throughput virtual screening and accurately evaluating vital photophysical properties of Pt-emitters.

The Future: Future research will explore more efficient representations of metal-complex materials and establish more advanced frameworks for predicting key properties

When developing Pt(II) complexes for OLED applications, the integration of HTVS-ML and Δ-learning stands out as a promising approach. The proposed robust protocol not only streamlines complexes generation and HTVS, but also delivers accurate predictions of key photophysical properties. Looking ahead, the continuous progress in artificial intelligence, ML, and quantum chemistry, particularly Δ-learning in materials science, holds immense promise for expediting materials discovery. Future endeavours will focus on addressing the challenges such as efficient representations of various materials and establishing advanced ML/deep learning frameworks for predicting properties of low-sample materials, where experimental and computational data are scarce.

The Impact: This work offers a promising pathway to achieving Pt complexes generation, HTVS and accurate predictions in phosphorescent OLED emitters and introduces innovative approach for solving low-sample issues in an ML way. 
The research has been recently published in the online edition of Materials Futures, a prominent international journal in the field of interdisciplinary materials science research.

Reference:
Shuai Wang, ChiYung Yam, LiHong Hu, Faan-Fung Hung, Shuguang Chen, ChiMing Che, GuanHua Chen. A general protocol for phosphorescent platinum(II) complexes: generation, high throughput virtual screening and highly accurate predictions[J]. Materials Futures. DOI: 10.1088/2752-5724/adb320

 

---

---