Dr Ronny Costi of ICL looks at how AI modelling can accelerate molecular discovery in the flame retardant field
Designing flame retardants for the low-loss electronics era: Man & machine in cooperation
Dr Ronny Costi of ICL looks at how AI modelling can accelerate molecular discovery in the flame retardant field
Every generation of electronics places new demands on the materials hidden in it. In the age of artificial intelligence( AI), high-speed communications, cloud computing, advanced servers, 5G and 6G systems, and high-frequency semiconductor packaging, the printed circuit board( PCB) is no longer merely a mechanical platform for copper conductors and components. It is a critical part of the signal path.
The PCB materials help determine how fast, cleanly and reliably electrical signals move through the system. They must therefore do more than survive heat, processing, and fire safety testing. They must support extremely fast signal transmission with minimal loss.
For decades, flame retardants( FRs) in printed PCBs were evaluated primarily by their ability to help a laminate to meet fire safety standards such as UL 94 V-0, while maintaining acceptable mechanical, thermal and processing properties. That framework is no longer sufficient. The new requirement is more demanding: FRs must protect against fire while remaining almost invisible to highspeed electrical signals.
This is a difficult materials challenge. In high-end copper clad laminates( CCLs), every resin component and additive can influence dielectric constant( Dk), which affects signal propagation speed and impedance control, and dissipation factor( Df), which governs dielectric loss, converting part of the signal energy into heat.
As data rates increase and trace geometries become tighter, even small increases in Df can become unacceptable. A FR that performs well in conventional electronics may therefore be unsuitable for ultra-low-loss and extreme-lowloss PCB systems.
ICL approached this challenge by seeking to develop a new generation of ultra-low-loss FRs for advanced PCB applications, including AI hardware, high-speed communications, servers and nextgeneration electronic infrastructure. The scientific hypothesis was clear: the right molecules would need to combine FR functionality with molecular features that minimise dielectric loss.
Finding those molecules through conventional laboratory screening alone would be too slow, too expensive, and too narrow. Instead, ICL partnered with NobleAI to apply computational chemistry, sciencebased AI( SBAI) and chemical AI modelling to the discovery process.
Why low-loss PCBs need new FRs
Low-loss PCB problem is not simply a computational exercise. It requires an understanding of fire chemistry,
68 SPECIALITY CHEMICALS MAGAZINE ESTABLISHED 1981