Bayesian Optimization-Assisted Screening to Identify Improved Reaction Conditions for Spiro-Dithiolane Synthesis.

Area of Science:

• Organic Synthesis
• Chemical Process Development
• Computational Chemistry

Background:

• 2,3,7,8-tetrathiaspiro[4.4]nonane (1) is a crucial intermediate for synthesizing tetramercaptan pentaerythritol.
• Efficient and scalable synthesis routes are essential for industrial applications.

Purpose of the Study:

• To optimize reaction conditions for the large-scale synthesis of compound 1 using Bayesian optimization.
• To enhance the yield and efficiency of the synthetic process.

Main Methods:

• Bayesian optimization (BO)-assisted screening was employed to identify optimal reaction parameters.
• Initial training sets (ITS) were generated using random screening and Latin hypercube sampling (LHS).
• The performance of BO was evaluated based on conversion yield within a limited number of experimental trials.

Main Results:

• An initial BO-assisted screening with random sampling achieved a 78% conversion yield of compound 1 within seven experiments.
• Subsequent BO-assisted screening utilizing an ITS selected by LHS further improved the yield to 89% within eight trials.
• The optimized conditions proved suitable for hundred-gram scale-up synthesis.

Conclusions:

• Bayesian optimization is an effective strategy for rapidly identifying improved reaction conditions in chemical synthesis.
• LHS-selected ITS enhances the efficiency of BO-assisted screening for process optimization.
• The developed methodology enables scalable and high-yield synthesis of key intermediate 1.