Grantee Research Project Results
Final Report: Machine-Learning-Assisted Development of Alternatives to Diarylide Pigments
EPA Contract Number: 68HERC22C0032Title: Machine-Learning-Assisted Development of Alternatives to Diarylide Pigments
Investigators: Saikin, Semion
Small Business: Kebotix, Inc.
EPA Contact: Richards, April
Phase: I
Project Period: December 1, 2021 through May 31, 2022
Project Amount: $99,910
RFA: Small Business Innovation Research (SBIR) Phase I (2022) RFA Text | Recipients Lists
Research Category: Small Business Innovation Research (SBIR) , SBIR - Toxic Chemicals
Description:
The goal of this proposed effort was to develop yellow pigments with synthetic pathways that do not unintentionally generate polychlorinated biphenyl (PCB) derivatives as by-products. PCBs are biphenyl compounds in which some (or all) of the hydrogen atoms are replaced by chlorine atoms and pose major complication with regards to human health and environmental toxicity. In particular, yellow diarylide pigments seem to produce the highest levels of PCBs due to the starting materials required in their production. Given that many products which use these diarylide yellows are formulated into consumer paints, plastics, etc., the generation of PCBs impacts a wide range of demographics and, thus, are strictly regulated. Therefore, there is high international demand from the continually growing pigment and dye industries to find alternative, sustainable materials. To aide in this demand, we employed our workflows which use ML methods for process control and optimization, deep-learning algorithms for prediction of molecular properties, and generative models for library development. From generated libraries, our semi-automated, high-throughput laboratory can perform sample preparation, synthesis, and characterization using workflows designed by expert chemists and engineers. Based off the best performing materials, toxicity screening criteria are integrated to evaluate which materials, and their associated synthetic by-products, provide cleaner, safer materials to work with at the industrial scale. Kebotix’s approach involves the use of a closed-loop platform, meaning each step or iteration of the pipeline is continually informed with data from a previous step/iteration. This approach allows our team to rapidly develop and bring alternative PCB-free yellow pigments to market, immediately reducing the unintentional production of PCBs and forcing long-term removal of probable carcinogens.
Summary/Accomplishments (Outputs/Outcomes):
At the end of phase I, we successfully deployed both of our proposed pipelines resulting in 7 promising candidates to replace PY13. Beginning with Pipeline 1, the first objective was to filter internal databases of synthesizable molecules. These structures include halogenated and/or toxic compounds as well as compounds with long lead times that would fall outside the timeframe of this project. Our initial task was to apply structural filters as well as available toxicity data to remove potentially hazardous compounds. The optical properties of these materials were filtered by predicted colorimetric values via ML models to the best possible candidates. The top candidates where then submitted for time-dependent density functional theory (TD-DFT) calculations for property verification, thus completing the second objective. After filtering, the top candidates were then acquired from commercial vendors. The candidates were then characterized in the lab for their colorimetric properties (i.e., reflectance spectra, CIE L*, C*, H* values) and then processed into a polypropylene (PP) matrix. For these studies, PP was chosen as the matrix given the benign interactions of the polyolefin with the pigment itself, allowing us to deduce the properties of the pigment when finely dispersed into plastic. Upon formation of the sample melts, leeching studies were performed, as well as the thermal and UV stability of the pigment:polymer samples. Figure 1 demonstrates samples discovered using our Pipeline 1 platform. Upon evaluation of the results, a common motif was established and used as an input in our proprietary retrosynthesis tool. Based off the outputs, a sustainable synthetic pathway was chosen, and a library of compounds were developed.
Figure 1. Photographic images of pigments candidates from the pipeline 1 in polypropylene identified through the pipeline before thermal/UV/leaching characterization. Pigment loading = 0.2 wt %.
Pipeline 2 began by procuring a list of feedstock molecules and generating a library of synthesizable candidates. Using the structural motif discovered during Pipeline 1, we sought to use a starting “core” and functionalize using a variety of aromatic substituents, thereby tuning the optical properties of the system. In this library, 11 candidates were designed, where 9 candidates were able to be isolated in purity levels above >90%. The candidates were dispersed into PP and polyamide-6 disks for characterization of colorimetric properties. Of these candidates, four compounds showed promising results with regards to heat stability and lightfastness in polypropylene and polyamide-6 disks, with one candidate showing good color characteristics as a PY13 replacement. With this success, we were inspired to use our pipeline for another synthetic library, this time focusing on retaining the industrial manufacturing processes used for pigment production. In this library, 18 candidates were designed, 15 of which were synthetically accessible (Table 1). These materials were characterized in a similar manner to the previous materials, of which three compounds showed very similar color characteristics to a standard PY13 sample when dispersed into a PP proxy matrix. Of the three materials, we employed our algorithms to identify the starting materials with the least predicted human and environmental toxicity.
Table 1. Proxy polypropylene disks of the PY13 library. Disk labeled Y13 is a reference disk which uses Pigment Yellow 13.
Conclusions:
Kebotix has successfully deployed their discovery pipelines, which uses ML, AI, and a semi-automated lab, towards the discovery, procurement, characterization, and evaluation of novel PCB-free colorants. At least two iterations from each pipeline were completed in the six-month timespan, thus highlighting its effectiveness and speed. A library of candidate molecules was developed, and 2 pigments were scaled up to 1 kg to provide samples to pigment producers and pigment users. Specifically, we have reached out to numerous ink and paint formulators. A handful of them are interested in sampling our PCB-free pigments. For faster market entry and adoption, we are currently in discussion with several pigment producers who are interested in either licensing the IP or being our manufacturing partners. With these commercial partners testing our pigments samples, we expect the materials will be commercialized in 2-3 years.
SBIR Phase II:
Machine-learning-assisted development of PCB-free alternatives to commodity pigmentsThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.