(662c) Automatic High Throughput Reaction Pathway Identification for More Sustainable Chemicals Production

The chemical industry is a known hard-to-abate industry due to its high dependence on fossil resources as sources of carbon-rich raw materials and energy for high-temperature industrial heating [1]. Unlike other sectors, such as the power sector where renewables can decouple electricity generation from carbon, the production of chemicals is intrinsically bound to carbon. This is due to the wide range of carbon-based chemical products (i.e., vastly different number of carbon atoms in the molecules) [1], and the dependence on high-temperature thermal energy [2]. To achieve a more sustainable chemicals production, we could transition towards a net zero emission (NZE) chemical industry by closing the so-called “carbon loop”. NZE pathways that present promising options for achieving a closed carbon loop include those that couple fossil feedstock technologies with carbon capture and storage (CCS), and others based on renewable resources, e.g., carbon capture and utilization (CCU), carbon recycling, and biomass [1].

Amongst these potential alternatives, CCU stands out as it utilizes captured CO₂, which is potentially emitted at the chemical’s production process as a direct emission and at the end-of-life phase via incineration or degradation. This CO₂ could replace the traditional carbon-rich fossil feedstocks, enabling an intrinsically circular production. CCU shows some advantages over other NZE pathways for the chemical sector, such as not relying on CO₂ underground storages and the avoidance of leakage risks, in contrast to CCS [3], and the lack of competition for the carbon feedstock [4] unlike biomass. According to many projections, it is likely that most, if not all, forms of NZE technologies will be implemented in some optimized combination in chemicals production, taking advantage of regional aspects, resource availability, and socio-political acceptance [5]. Despite the recognized potential of CCU as a necessary part of achieving a sustainable chemical industry [5, 6], this pathway still faces the challenges of having a high energy and electricity demand [7] and, critically, many of the CO₂-based production processes are still at low technology readiness level (TRL), only being proven in the lab scale [8]. Although many researchers have made great progress in the study of reaction feasibility, reaction mechanisms, catalysts, etc., there are still large gaps in the understanding of what potential these new technologies have for profitably reducing the reliance on fossil resources and CO₂ emissions [5, 8].

At present, experimental groups develop new potentially appealing routes, whose sustainability performance is seldom investigated in-depth in the original publications. Based on such articles, modeling groups develop techno-economic and environmental studies, often focusing on the most promising or mature technologies. This decoupled approach implies that the discovery of promising new sustainable processes relies often on an iterative, and manual method involving chemical intuition, simulations, and experimentation – often costly and time-consuming. Alternatively, we could employ chemical reaction databases such as Reaxys, CAS SciFinder, and USPTO to identify new promising chemical reaction pathways from sustainable feedstocks , which would then be analyzed more in-depth using process simulation and LCA.

A number of works have already utilized the said chemistry datasets for different purposes, including creating and analyzing the network of organic chemistry [10] (NOC), which could enable the rapid search for innovative reaction pathways or key chemicals necessary to the NZE transition in the chemical industry. The vast application of the NOC includes statistical network analyses [11], identification of strategic molecules for future circular supply chains [12], and investigations on the cyclic nature of organic chemistry [13]. Additionally, the NOC allows the identification of sustainable multistep chemical reaction pathways for fine chemicals within simplified small sub-networks [14]. Based on these datasets, large-scale screening of new reaction pathways can be completed either manually combined with human chemical intuition [8] or programmatically via high throughput screening to drive the direction of more extensive research [9]. The main limitation of these approaches is the availability of data. Manual exploration approaches are typically bounded by the intuition and knowledge of the human users. Alternatively, programmatic exploration is severely hindered by the prevalence of data gaps and unprocessed data.

To contribute towards the automatic exploration and identification of promising sustainable chemical routes for bulk and fine chemicals production, we introduce a high throughput reaction pathway screening tool based on the Reaxys chemistry database. This work innovates on similar works by introducing methods to automatically fill in critical data gaps, such as stoichiometry, and a greedy search algorithm to search the NOC for promising sustainable reaction pathways. In this exemplary case applied to CCU, the high throughput tool generates a localized bipartite graph network representation of the NOC by querying the Reaxys API. The gathered data is processed and any critical data gaps are filled using optimization or other property estimation techniques. Incomplete stoichiometric reactions are completed using optimization models that propose chemically valid completed stoichiometric equations. Green chemistry-inspired key performance indicators (KPIs), such as atom economy, hydrogen efficiency, etc. are calculated for each reaction. To traverse the dense graph from CO₂ to a desired target bulk or fine chemical, a greedy, depth-first, graph path search algorithm is developed. The algorithm explores along the best-performing reaction pathway until it reaches the target chemical and restarts, always searching along the next best partially explored or unexplored reaction pathway until the desired number of reaction pathways have been collected.

Based on the KPI selected, the high throughput screening tool delivers a ranked list of the most promising full reaction pathways, i.e. pathways connecting CO₂ and other selected cutoff raw materials to the target chemical. The ranked list can be further characterized by the type of reactions occurring within the reaction pathways, e.g. electrochemical or thermocatalytic reactions, that can be specifically analyzed or independently compared. With this work, we enable the broader audience of researchers investigating sustainable chemistry and in particular, CCU, to have access to relevant and curated data so that they might identify and focus efforts on, by design, promisingly sustainable reaction pathways for our transition to a future NZE chemical industry. Our results show that the method proposed can identify pathways performing well in a range of sustainability metrics that would be very hard to find using simple iterative methods based on heuristics.

References

[1] P. Gabrielli et al., “Net-zero emissions chemical industry in a world of limited resources,” One Earth, vol. 6, no. 6, pp. 682–704, Jun. 2023, doi: 10.1016/j.oneear.2023.05.006.

[2] C. Chung, J. Kim, B. K. Sovacool, S. Griffiths, M. Bazilian, and M. Yang, “Decarbonizing the chemical industry: A systematic review of sociotechnical systems, technological innovations, and policy options,” Energy Research & Social Science, vol. 96, p. 102955, Feb. 2023, doi: 10.1016/j.erss.2023.102955.

[3] E. A. Quadrelli, G. Centi, J. Duplan, and S. Perathoner, “Carbon Dioxide Recycling: Emerging Large‐Scale Technologies with Industrial Potential,” ChemSusChem, vol. 4, no. 9, pp. 1194–1215, Sep. 2011, doi: 10.1002/cssc.201100473.

[4] A. Muscat, E. M. De Olde, I. J. M. De Boer, and R. Ripoll-Bosch, “The battle for biomass: A systematic review of food-feed-fuel competition,” Global Food Security, vol. 25, p. 100330, Jun. 2020, doi: 10.1016/j.gfs.2019.100330.

[5] P. Markewitz et al., “Worldwide innovations in the development of carbon capture technologies and the utilization of CO2,” Energy Environ. Sci., vol. 5, no. 6, p. 7281, 2012, doi: 10.1039/c2ee03403d.

[6] Á. Galán-Martín, V. Tulus, I. Díaz, C. Pozo, J. Pérez-Ramírez, and G. Guillén-Gosálbez, “Sustainability footprints of a renewable carbon transition for the petrochemical sector within planetary boundaries,” One Earth, vol. 4, no. 4, pp. 565–583, Apr. 2021, doi: 10.1016/j.oneear.2021.04.001.

[7] I. Ioannou, Á. Galán-Martín, J. Pérez-Ramírez, and G. Guillén-Gosálbez, “Trade-offs between Sustainable Development Goals in carbon capture and utilisation,” Energy & Environmental Science, vol. 16, no. 1, pp. 113–124, 2023, doi: 10.1039/D2EE01153K.

[8] A. Otto, T. Grube, S. Schiebahn, and D. Stolten, “Closing the loop: captured CO 2 as a feedstock in the chemical industry,” Energy & Environmental Science, vol. 8, no. 11, pp. 3283–3297, 2015, doi: 10.1039/C5EE02591E.

[9] J. M. Weber, Z. Guo, C. Zhang, A. M. Schweidtmann, and A. A. Lapkin, “Chemical data intelligence for sustainable chemistry,” Chem. Soc. Rev., vol. 50, no. 21, pp. 12013–12036, 2021, doi: 10.1039/D1CS00477H.

[10] C. M. Gothard et al., “Rewiring Chemistry: Algorithmic Discovery and Experimental Validation of One‐Pot Reactions in the Network of Organic Chemistry,” Angewandte Chemie, vol. 124, no. 32, pp. 8046–8051, Aug. 2012, doi: 10.1002/ange.201202155.

[11] P.-M. Jacob and A. Lapkin, “Statistics of the network of organic chemistry,” React. Chem. Eng., vol. 3, no. 1, pp. 102–118, 2018, doi: 10.1039/C7RE00129K.

[12] J. M. Weber, P. Lió, and A. A. Lapkin, “Identification of strategic molecules for future circular supply chains using large reaction networks,” React. Chem. Eng., vol. 4, no. 11, pp. 1969–1981, 2019, doi: 10.1039/C9RE00213H.

[13] M. D. Bajczyk, P. Dittwald, A. Wołos, S. Szymkuć, and B. A. Grzybowski, “Discovery and Enumeration of Organic‐Chemical and Biomimetic Reaction Cycles within the Network of Chemistry,” Angew Chem Int Ed, vol. 57, no. 9, pp. 2367–2371, Feb. 2018, doi: 10.1002/anie.201712052.

[14] A. A. Lapkin et al., “Automation of route identification and optimisation based on data-mining and chemical intuition,” Faraday Discuss., vol. 202, pp. 483–496, 2017, doi: 10.1039/C7FD00073A.