Linking Molecular Structure via Functional Group to Chemical Literature for Establishing a Reaction Lineage for Application to Alternatives Assessment

Abstract

The evaluation of potential alternatives for chemicals of concern (CoC) requires an understanding of their potential human health and environmental impacts during the manufacture, use, recycle and disposal life stages. During the manufacturing phase, the processes used to produce a desired chemical are defined based on the sequence of chemical reactions and unit operations required to produce the molecule and separate it from other materials used or produced during its manufacture. This paper introduces and demonstrates a tool that links a chemical’s structure to information about its synthesis route and the manufacturing process for that chemical. The structure of the chemical is entered using either a SMILES string or the molecule MOL file, and the molecule is searched to identify functional groups present. Based on those functional groups present, the respective named reactions that can be used in its synthesis routes are identified. This information can be used to identify input and output materials for each named reaction, along with reaction conditions, solvents, and catalysts that participate in the reaction. Additionally, the reaction database contains links to internet references and appropriate reaction-specific keywords, further increasing its comprehensiveness. The tool is designed to facilitate the cataloging and use of the chemical literature in a way that allows user to identify and evaluate information about the reactions, such as alternative solvents, catalysts, reaction conditions and other reaction products which enable the comparison of various reaction pathways for the manufacture of the subject chemical. The chemical manufacturing processing steps can be linked to a chemical process ontology to estimate releases and exposures occurring during the manufacturing phase of a chemical.

Graphical Abstract

Introduction

The Toxic Substances Control Act (TSCA) of 1976 (P.L. 94–469) provided the US Environmental Protection Agency (US EPA) with the nation’s primary chemicals management law (15 U.S.C. §2601 et Seq.). The act furnished US EPA the authority to require reporting, record-keeping, testing, and restrictions relating to chemical substances and/or mixtures. Forty years later, TSCA was amended by the Frank R. Lautenberg Chemical Safety for the 21st Century Act (P.L. 114–182), which included much needed improvements such as a mandatory requirement for US EPA to prioritize and conduct risk-based reviews of new and existing chemicals with clear and enforceable deadlines, increased transparency for chemical information, and a consistent source of funding for US EPA to carry out its responsibilities under the new law.¹ Another facet of the amended TSCA is that chemical risk evaluations will be state-of-the-science and consider risks associated with the chemical across the full life cycle of a chemical, including potential exposures associated with occupational health and safety risks.²

During the chemical risk assessments conducted under TSCA, a chemical may be identified as chemicals of concern (CoC), which initiates the identification and assessment of alternatives that serve as potential replacements for the CoC. A central component of Alternatives Assessment (AA) is the application of a life cycle approach to evaluating the safety of a CoC and its alternatives. In 2014, the National Research Council of the National Academy of Sciences (NAS) produced “A Framework to Guide Assessment of Chemical Alternatives,”³ which describes a 13-step framework, which aids in a decision-making process when assessing alternatives to a CoC. A key component of chemical safety evaluation is the estimation of cumulative chemical exposures during the manufacturing, use, recycling and disposal stages of a chemical’s life cycle. Chemical releases during the manufacturing of a chemical or its formulation within a product results in occupational exposures to workers, while emissions from the manufacturing process results in exposures to the public. In addition, the effects of near-field exposures to consumer products can be incorporated into the alternatives assessment for the chemical.⁴

The US EPA has developed “top-down”⁵ and “bottom-up”⁶ approaches to estimate life cycle inventories (LCIs) and mass and energy flows, for chemical processes attributed to the manufacturing phase of the chemical’s life cycle. The “top-down” method involves mining facility-level release data from US EPA databases such as the Toxics Release Inventory (TRI),⁷ the National Emissions Inventory (NEI)⁸ and the Facility Registry Service (FRS).⁹ The “bottom-up” method estimates LCIs using computer simulations and detailed process flow diagram calculations. US EPA has developed a methodology to reconcile the results of the “top-down” and “bottom-up” approaches, which can validate each of these methods, as well as hybridize estimation methods to meet the needs of the decision-maker.¹⁰ To expand upon these approaches for the estimation of LCI, US EPA has developed a lineage ontology that provides a general framework to conceptualize the synthesis steps required to produce a chemical from raw materials or elementary flows, such as crude oil or biomaterials, and the process information to manage data describing the various unit processes associated with each synthesis/process step.¹¹

To populate these ontologies, which can be used to generate LCIs, specific synthesis information for a chemical under evaluation must be provided. Therefore, US EPA has introduced framework for sustainable chemical synthesis design to guide more sustainable chemical synthesis using molecular structure information coupled with a reaction synthesis database.¹² This paper presents a methodology and computational structure for implementing the Sustainable Chemistry Synthesis Expert Framework as a multi-level software tool for the systematic use of the sustainable chemical synthesis design framework that enables users to identify sustainable chemical synthesis routes. The expert framework tool links a chemical’s functional groups to named reactions to provide information regarding its synthesis route, and information for determining the manufacturing process for that chemical. This synthesis information can be used to populate the reaction ontology and can contribute information about the process ontology to aid in the automation of LCI generation.

Assessment of Alternative Chemical Manufacturing Processes

As the importance of attaining more sustainable chemical manufacturing processes continues to increase, the need to capture all the necessary and relevant information about the manufacturing process becomes a priority. Additionally, the ability to incorporate the synthesis and process conditions aids in defining and assessing alternative chemical synthesis/process routes, which truly leads to a comprehensive assessment of the entirety of the manufacturing phase of the life cycle. This robust analysis allows for a more accurate and respective comparison of a CoC and its alternatives.³

Therefore, comprehensive alternatives assessment requires the identification of all the chemicals involved in the entire synthesis route used to manufacture the CoC being assessed. The chemical lineage ontology conceived by the US EPA scientists¹¹ identifies the parent compounds for a desired chemical product (child). Building a chemical lineage facilitates the identification of the LCI for the chemical by including LCI information of not only the parents of the desired chemical, but also the grandparents and earlier ancestors in the chemical’s lineage, tracking back to elementary flows (e.g. crude oil, coal, natural gas, water). The LCI of the desired chemical could then be obtained by aggregating the LCI information of the chemicals included in the chemical’s lineage. Reactants (parents), product (child), and by-products (siblings) from each lineage level involved in a synthesis route might pose a greater environmental impact than the desired product. Mittal, et. al. (2018)¹¹ determined the chemical lineage of caprolactam, which is used in the manufacture of Nylon 6, to facilitate the compilation of its life cycle inventory. Caprolactam’s chemical lineage includes benzene, toluene, cumene, carbolic acid, and other intermediates produced along the lineage from fossil resources (i.e. coal, natural gas, or petroleum) to the caprolactam product. The LCI for caprolactam was then assembled from the life cycle inventories of these lineage chemicals. This ontological approach enables for an expanded assessment that can aid in the identification of different synthesis routes for the same final product, to determine if an alternative synthesis route can be utilized that includes less toxic reactants, prevents the generation of undesired by-products, or generates less waste and uses less energy. The results of the Sustainable Chemistry Synthesis Expert Framework can be used to identify synthesis routes for a given chemical and construct the lineage of the product chemical.

This previously introduced reaction ontological data can be used within a two-stage methodology for comparing chemical alternatives.¹³ In the first stage, hazard scores for the alternatives are evaluated for a variety of human health outcomes, acute mammalian toxicity (via the oral, dermal, and inhalation routes), carcinogenicity, mutagenicity, endocrine disruption, reproductive and developmental toxicity, neurotoxicity, systemic toxicity, eye and skin irritation, and skin sensitization), ecotoxicity (acute and chronic aquatic toxicity), and physicochemical properties (persistence and bioaccumulation). In the second stage, the identified alternatives are evaluated by combining the hazard scores with exposure to the chemical to evaluate the chemical alternatives in terms of risk over the entire life cycle.

To facilitate the implementation of the first stage of this methodology, a new online Alternatives Assessment Dashboard (AA Dashboard) is currently being developed. The AA Dashboard will provide information that can be used to compare chemical alternatives. The AA Dashboard will generate hazard scores based on data from quantitative experimental toxicity values, global harmonization system (GHS) scores, quantitative structure activity relationship (QSAR) models, and government lists/databases. For each endpoint, data from the multiple sources will be converted into qualitative scores (low, medium, high, and very high) that represents the hazard profile for the chemical. Scores will be color-coded to highlight categories of the greatest concern, which aids in the identification and selection of the safer or greener alternative. The on-line AA Dashboard will allow users to enter the identified alternatives to be compared by searching for chemicals in the US EPA’s Chemistry Dashboard¹⁴ and clicking on the calculate button to generate a table that lists the hazard profile for each proposed alternative. The reaction ontology can be used to suggest additional chemicals involved in the synthesis route to gain a more comprehensive comparison of chemical alternatives.

To implement the second stage of the methodology, exposure values can be estimated using models such as the Stochastic Human Exposure and Dose Simulation Model for Multimedia, Multipathway Chemicals High Throughput (SHEDS-HT),¹⁵ which was developed under the ExpoCast program.¹⁶ The second stage enables a more accurate comparison of potential exposures to each alternative and the consideration of additional factors that may not be obvious based on separate binned persistence, bioaccumulation, and toxicity scores. Risk can be evaluated in terms of the hazard-to-exposure ratio (HER).

Additionally, a module will be added into the on-line AA Dashboard to display the environmental breakdown products from US EPA’s Chemical Transformation Simulator (CTS).^17–18 To further expand on our efforts, US EPA is developing an on-line tool, RapidTox, to integrate available chemistry, toxicity and exposure information to help decision-makers to quickly and efficiently evaluate chemicals.^19–20 Examples of RapidTox applications include pre-prioritization of chemicals under TSCA and developing screening level toxicity values to aid in planning for Superfund site cleanup.²¹ RapidTox will be able to generate qualitative scores (low, medium, and high) for an array of categories, like the on-line AA Dashboard. These scores are within the domains of hazard, exposure, and persistence/bioaccumulation. RapidTox may include methods to calculate an overall score across within these domains. Data and models developed for the on-line AA Dashboard can be utilized to supplement RapidTox data. Additionally, the web-based tools developed for the on-line AA Dashboard (e.g. the hazard profile comparison table) can be used in RapidTox to provide decision-makers with more detailed information for comparing chemicals. Each of these features will expand the comprehensiveness of information available for decision-makers who are evaluating chemical alternatives for replacement or for the prioritization of chemicals for additional testing.

Overview of the Sustainable Chemistry Framework

As described in the previous contribution,¹² the approach utilized by the Sustainable Chemistry Synthesis Expert Framework identifies all of the functional groups present in a molecule and labels them as reaction centers. By employing this “deconstruction” approach, the identified functional groups can be coupled with organic chemistry synthesis rules to provide a schematic framework akin to retrosynthesis. This information allows for sequencing and selection of needed name reactions for the molecule’s generation. This approach is designed to mimic how an organic chemist envisions a molecule’s synthesis upon first sight. These reaction centers may be identified through various mechanisms, including maximum common substructure (MCS) or extended connectivity algorithms.²² In this paper, we identify a method for the identification and qualification of the reaction centers present based on the functional groups in the molecule. Based on the type of functional group(s) present in the molecule, potential named chemical reactions that could be used in their synthesis and result in the formation of that compound are provided as an output by the tool. This paper presents a case study demonstrating the detection of different reactions that can be used in the manufacture of a phosphorus-based flame retardant, dimethyl methylphosphonate, which is a proposed replacement for brominated flame retardants (BFRs). Named reactions for the manufacture of dimethyl methylphosphonate will be identified so that a comprehensive alternatives assessment can be performed to assess the toxicity and various sustainability metrics for the reactions. Additionally, the chemical lineage for each reaction can be identified through the functional groups of compounds that participate in the named reactions. This chemical lineage can then be tied to the known life cycle inventories for ancestor compounds.

This tool utilizes a database of functional groups and reactions that have been compiled manually from the literature and provides a mechanism to understand the chemical releases that may occur as a part of the manufacturing process, through identifying potential named chemical reactions used to manufacture the chemical. Knowledge of the chemical reaction used to manufacture a chemical provides information on the reaction conditions (temperature, pressure, reactants, etc.) and reaction products that drive the overall processing steps (e.g., separations) required to produce the desired chemical. This information can then be used to propose more benign alternative starting materials, thus helping to identify more sustainable chemical synthesis routes.

The database can be enhanced to include additional variations of the named reactions, and how the modifications affect the sustainability metrics associated with the identified reaction. Using this information, chemists can evaluate the sustainability of the reaction as part of the overall manufacturing process. As reactions are added to the database, machine learning technologies can be applied to help evaluate the sustainability of candidate reactions and manufacturing processes.

System Design

A block diagram illustrating the main components of the Sustainable Chemistry Synthesis Expert Framework is shown in Figure 1.

Figure 1.

Block diagram showing components of the Sustainable Chemical Synthesis Framework.

The tool consists of a chemical reaction database and a functional group database. The current application was written in C#, but can be exported as a web service and accessed using Python or JavaScript. Source code for the application is available at https://github.com/USEPA/sustainable-chemistry-synthesis-expert-framework. In its current form, the tool consists of a chemical informatics class library and Windows Forms desktop application for use on Microsoft Windows-based computers. The chemical informatics library includes a SMILES parser written using the ANother Tool for Language Recognition (ANTLR) Version 4 parser generator (http://www.antlr.org) from a grammar based on the OpenSmiles specification (http://opensmiles.org/opensmiles.html). The functional group database will contain the name of the functional group, its associated structure in the form of a SMiles ARbitrary Target Specification (SMARTS) string, and an image file stored as a Graphics Interchange Format (gif). Selected phosphorous functional groups, their SMARTS structures, and functional group image are provided in Table 1. The entire functional group database contains 219 functional groups and is presented as an Excel spreadsheet in the supplemental information.

Table 1.

Examples of Functional Groups Containing Phosphorous, Carbon and Oxygen from the Functional Group Database.

Functional Group	SMARTS	Image
Phosphate	OP(=O)(O)O
Phosphinate	OP(=O)(C)C
Phosphine Oxide	CP(=O)(C)C
Phosphinite	CP(C)C
Phosphite	OP(C)O
Phosphonate	OP(=O)(O)C
Phosphonite	OP(O)C
Phosphorane	CP(C)(C)(C)C

The reaction database contains data regarding the reaction including the functional groups of parent compounds that participate as reactants, and the functional groups of the resulting product (child) and/or by-products (siblings) of the reaction. In addition, the reaction database contains data on the reaction conditions, solvents, catalyst types, and other information relevant to the reaction. From this information, the reaction database can be used to provide information on environmental impacts of the reactants, products, and process conditions used in the synthesis of the desired final product. This is compiled to enable users to identify more sustainable synthetic methods. Additionally, the application provides links to chemical literature related to the reactions or chemicals of interest contained in external databases, increasing the comprehensiveness and ability for the tool to have up to date information, in the form of journal articles and subscription services. Table 2 provides selected information from the Reaction database for the phosphorous compounds used in the case study. The full reaction database contains data for 181 reactions is included as an Excel spreadsheet in the Supplemental Information.

Table 2.

Selected Database Entries for the Reactions used in the Production of Phosphorous-based Compounds.

Functional Group Formed	Named Reaction	Reaction Image	Reactant Functional Groups		Acid Base	Heat	Catalyst	Solvent
Phosphinate	Dehydrogenative Coupling		Phosphine oxide	Alcohol	K2CO3	130°C	Iron	Toluene
	Alkylation		Phosphinic acid	Alkyl halide	K2CO3 TEA	115°C	Base	Acetonitrile
	Hirao reaction		H-phosphinate	Bromobenzene	K2CO3	150°C	NiCl2	Acetonitrile
	Oxidation		Phosphine	--	None	0°C	H2O2	Water
Phosphine	Selective Reduction		Phosphine Oxide	--	None	100°C	Silanes	Toluene
	Selective Reduction		Phosphine sulfide	Halobenzene	None	Ambient	Nickel	Benzene
	Selective Reduction		Phosphorus chloride	Halobenzene	None	200°C	AlCl3	Halobenzene
	Selective Reduction		Phosphorous	Halobenzene	None	20°C	Titanium reagent	Benzene
Phosphine Oxide	Phosphine Oxidation		Phosphine	Oxygen	None	20°C	NaBrO3	Water
	Phosphine Oxidation		Phosphine sulfide	--	None	20°C	PhIO2	Acetonitrile
	Phosphine Oxidation		H-phosphinate	Sulfonyl dibenzene	NaOBut	20°C	Nickel	Dioxane
Phosphinite	Reduction		Chlorophosphine	Methanol	TEA	20°C	TEA	Ether
Phosphite	Phosphite Production		Phosphorus trichloride	Methanol	Ammonia Diethylamine	20°C	Base	Xylene
Phosphonate	Michaelis-Abruzov		Phosphite	Halide, Alkyl	None	Reflux	None	Toluene
	TMS-promoted Michaelis-Abruzov		Phosphite	Tms-Br	None	100°C	None	None
	Alkylphosphonic acid esterification		Phosphonic Acid	Alcohol	None	Room temperature	p-TsOH or silica chloride	Methanol
	Solid Phase		Methylphosphonic dichloride	Alcohol	None	Room temperature	Neutral alumina	None

The chemical informatics package consists of classes for atoms, and chemical bonds that are used as part of the molecule class designation, as shown in Figure 2. The molecule class is based on graph-theory, and atoms represent the nodes of the graph, and chemical bonds are the edges between the nodes (atoms) present in the molecule being evaluated. Atom classes contain structural information such as details on adjacent atoms and the chemical bonds that join them together. The chemical informatics class library implements graph theory algorithms, including subgraph isomorphism and ring finding protocols. The subgraph isomorphism algorithms include Ullman²³ and the VF2 algorithm.²⁴ All ring finding sub-structure surveys will be performed using Tarjan’s depth first search algorithm.²⁵

Figure 2.

Class diagram for the Sustainable Chemistry Synthesis Framework.

Currently, a desktop application has been developed for use on local machines, primarily for testing of the chemical informatics class library. This application allows identification of potential reactions to form the desired chemical compound (product and child). The chemical informatics library will be accessed via a web service. The web service output will be either a JavaScript Object Notation (json) string or a html-formatted string. The json output functionality allows for its use by external applications. The html output can be directly displayed in a webpage, such as on EPA’s National Center for Computational Toxicology’s (NCCT’s) Chemistry Dashboard (https://comptox.epa.gov/dashboard/).

Future development of the application may incorporate computer-aided organic synthesis (CAOS) to utilize machine learning approaches that allow limited chemical reaction identification and retrosynthesis planning.^26–28 Yadav²⁹ states the limitations of the application of machine learning approaches to the design of organic synthesis routes include the number of plausible synthetic routes, the size of the reaction databases required, and inconsistencies within the reaction database schemas. Therefore, we intend to utilize crowd sourcing of the development and population of the functional group, named reactions, and literature reference databases.

Greenness evaluation of the Synthesis Processes

Since the introduction of the twelve Green Chemistry principles in 1998,³⁰ various types of metrics have been developed for the evaluation of the greenness of a chemical process.^31–34 The twelve principles of green chemistry, which can be implemented in every day chemical processes, include atom economy (AE),³⁵ waste prevention, less hazardous chemical synthesis and use, energy efficiency, safer solvents and auxiliaries, renewable feedstocks, catalysis, reduced derivatives, real-time analysis for pollution prevention, and inherently safer chemistry for accident prevention. The main emphasis of green chemistry is to reduce or eliminate environmental impacts by reducing or eliminating the use and generation of hazardous materials in the design, manufacturing, and application of chemical products.

Sustainable chemical process development utilizes the principles of green chemistry and incorporates the twelve principles of Green Engineering,³⁶ as well as economic and social impact evaluations. The simplest green chemistry metric is AE, which is the ratio of the mass of the reactants that is incorporated into the final desired product. As AE increases, a larger percentage of the mass of the reactants end up in the desired product, as opposed in contrast to being incorporated into undesired byproducts that may be disposed as waste. This metric can be easily calculated using stoichiometric information for that reaction. The AE only considers the mass of reactants incorporated into the product, and ignores solvents, catalysts, or the auxiliaries in its calculation. Sheldon³⁷ introduced the E-factor, which is the ratio of the mass of waste to the mass of product. The E-factor defines waste as all materials required for the reaction that is not the product, including all solvents, spent catalysts or other reaction materials. Since the E-factor does not differentiate between toxic waste and benign waste, a process producing a large amount of a benign salt by-product is considered worse when compared to a process where a small quantity of toxic by-product is produced. To overcome this shortcoming, the Effective Mass Yield (EMY), which is the percentage of the mass of the product relative to the mass of all non-benign materials used in the synthesis, was introduced.³⁸

The American Chemical Society Green Chemistry Institute Pharmaceutical Round Table (ACS GCI-PR) proposed the Process Mass Intensity (PMI),³⁹ which is the ratio of the quantity of raw materials input to the quantity of bulk product output. The PMI takes into consideration all the materials used in the process, including all the solvents used for the reaction as well as for the isolation and purification of the product, and the yield of the product. Due to the availability of these various metrics to evaluate the greenness of a process, it is difficult to identify a metric that is best-suited for all processes. Therefore, it is appropriate to identify a metric that applies to a specific process depending on the progress (stage) of a process. For example, when the process is at the beginning of the research stage, the AE can help to identify a synthesis path that provides the highest theoretical yield. In the later stages of the process development of a chemical synthesis process, other metrics including the E-Factor, EMI and PMI can be used. It is particularly critical to use these latter green metrics to evaluate the greenness of a commercial-scale chemical process.

It is evident that in most chemical processes solvents are used in a large excess when compared to reactants, reagents, and products. It is estimated that approximately 50–70% of the materials used in a pharmaceutical API production process are solvents,⁴⁰ and the majority of these solvents pose an environmental concern. Thus, an ideal green reaction is a reaction for which the solvent has been eliminated, with the next choice being a reaction in which a minimal amount of a greener solvent is used.

Results and Discussion

To demonstrate the utility of this tool, a case study is presented using proposed alternatives to the class of commercially-produced brominated flame retardants (BFRs). There are three main classes of brominated flame retardants, 1. polybrominated diphenyl ethers (PBDEs), 2. hexabromocyclododecanes (HBCDs) and 3. tetrabromobisphenol A (TBBP-A). These chemicals are added to various classes of materials such as textiles, building materials, and electronic equipment to suppress combustion and delay the spread of flame. Due to their perseverance in the environment, ability to bioaccumulate and potential toxicity to humans, certain BFRs, specifically penta-BDE, Octa-BDE and deca-BDE, are banned by the European Union.⁴¹ In 2004, some US manufacturers voluntarily halted the production of penta- and octa-BDEs. Due to the ban and the decision to not manufacture these brominated flame retardants, organophosphates, which are already being used as flame retardants, are considered to be possible alternatives for BFRs.⁴² When alternatives are chosen, these new chemicals must function as efficiently and have less toxicity, lower environmental impacts, and improved biodegradability than the chemicals they are replacing. It is critical to evaluate the potential alternatives using life cycle assessment (LCA) methods to qualify and quantify the potential environmental and human health impacts associated with all stages of its production, use and disposal. Thus, an initial aspect to be considered in a LCA is the manufacturing process itself.

Dimethyl methylphosphonate was chosen as an alternative to BFRs as an example to demonstrate the framework’s ability to elucidate the synthesis processes that are readily available in the literature and to evaluate their potential as more sustainable processes. The tool identified two functional groups, the phosphonate functional group, and the phosphonate ester functional group in the dimethyl methylphosphonate molecule. The framework tool then found four different named reactions from the reaction database that could be used for the synthesis of the phosphonate ester functional group. The first two schemes identified are based on Michaelis-Arbuzov rearrangement. The third method is the direct esterification of methylphosphonic acid. The fourth method for dimethyl methylphosphonate preparation involves the reaction of methylphosphonic dichloride with an alcohol in the presence of a base.

Figure 3 provides the output of the tool showing the reaction and literature available for producing dimethyl methylphosphonate via the Michaelis-Arbuzov rearrangement of phosphites that utilizes 1 equivalent of trimethyl phosphite and 1 equivalent of methyl iodide. The reaction is performed in toluene under reflux temperature for about 6 hours.^43–44 The AE for this reaction is approximately 47% and 1 equivalent of methyl iodide is generated as a by-product for each equivalent of product generated. This reaction has been improved by applying microwave energy in the absence of solvent.⁴⁵ Under microwave irradiation conditions, the reaction is completed in 5 minutes and the product is isolated with a 95% yield. Thus, applying alternate microwave energy improved the process by reducing the reaction time and energy, as well as eliminating the need for a solvent. Although the overall process greenness has improved, the AE for this reaction is still at 47%.

Figure 3.

Application output showing Michaelis-Arbuzov Rearrangement of Phosphite to Phosphonate.

The reaction efficiency for this Michaelis-Arbuzov reaction is further improved by using a catalytic amount (5%) of trimethylsilyl-bromide (TMS-Br)⁴⁶ instead of using one equivalent of methyl iodide (Figure 4) as well as generating an equivalent of methyl iodide, which improved the AE from 47% to 94%. This reaction is conducted with trimethyl phosphite and 5% trimethylsilyl bromide at 100 °C for 3 hours to yield 91% of dimethyl methylphosphonate. This Arbuzov reaction was successfully performed using catalytic amounts of iodine (0.1 mol% iodine, reflux for 24 hours, yield 99%)⁴⁷, or with a Lewis acid (5 mol% TMSOTf, chloroform, 60 °C, 18 h, 98% yield)⁴⁸ instead of TMS-Br.

Figure 4.

Application output showing Michaelis-Arbuzov Rearrangement of Phosphite using TMS-Br to Phosphonate.

The third reaction, shown in Figure 5, utilizes the esterification of 1 equivalent of methylphosphonic acid with 2 equivalents of methanol in the presence of an acidic catalyst, such as immobilized p-TsOH on Celite,⁴⁹ silica chloride,⁵⁰ and polymer-bound triphenylphosphine and iodine.⁵¹ The AE of this route is 77.5% with water being the only by-product generated. The esterification of the methylphosphonic acid with methanol in the presence of silica chloride is accomplished in 20 minutes at room temperature and provides an 89% yield of the desired product. In a similar way, esterification of methylphosphonic acid with methanol at room temperature for 15 minutes in the presence of immobilized p-TsOH on Celite provided dimethyl methylphosphonate at an 87% yield. Although the polymer-bound triphenylphosphine reaction produces dimethyl methylphosphonate at an 85% yield, this methodology uses 2 equivalents of iodine and imidazole as reactants, which will reduce the AE to 77.5% and produces 2 equivalents of imidazolium iodide waste. Additionally, this reaction uses volatile and hazardous dichloromethane as the solvent.

Figure 5.

Application output showing the Esterification of Alkylphosphonic acid.

Figure 6 shows the reaction of methylphosphonic dichloride with methanol, which was performed for 20 minutes at ambient temperature in the presence of neutral alumina in the absence of solvent, and the product was isolated at a 95% yield.⁵² Again, with this exchange reaction two equivalents of HCl are generated for each equivalent of product formed.

Figure 6.

Application Output Showing the Reaction of Methylphosphonic Dichloride with Methanol.

When the AE for each of these reactions is compared against one another, the AE for the Arbuzov reaction, shown in Figure 5, is the highest at 94% and the dehydration reaction, shown in Figure 6, is the second highest with 77.5%. When we compare these two methods and their experimental conditions, the AE for the Arbuzov reaction is slightly higher compared to the dehydration reaction, with the yield of the product being similar in both processes. The Arbuzov reaction requires higher energy and longer reaction times for rearrangement to occur, whereas the dehydration reaction occurs at ambient temperature with shorter reaction times. However, dehydration reactions require other catalysts to facilitate the reaction. It is possible to reduce the Arbuzov (Figure 5) reaction time and energy requirement by utilizing microwave energy as demonstrated in one of the studies that reaction time can be reduced to 5 minutes or alternatively a continuous flow system can be used. In addition to these parameters, it is important to consider the economic factors of the process parameters including the cost of the raw materials (trimethyl phosphite is less expensive than methylphosphonic acid), reagents and catalysts and the societal impacts to identify a sustainable process.

The hazard profiles from the beta version of the AA Dashboard for chemicals involved in the synthesis of dimethyl methylphosphonate are provided in Table 3. Methyl phosphonic dichloride is estimated to be more acutely toxic than trimethyl phosphite and methyl phosphonic acid. Thus, methyl phosphonic dichloride may not be the preferred phosphate starting material. Methyl iodide has a very high aquatic toxicity score and therefore may not be the best catalyst option. Similarly, bromotrimethylsilane has a very high eye and skin irritation score. Methyl phosphonic acid and methanol may be the least toxic option for producing dimethyl methylphosphonate, however, the lack of scores for methyl phosphonic acid for most toxicity categories may indicate data gaps within the beta version of the AA Dashboard.

Table 3.

Hazard profiles from the Beta Version of the beta-version of the Alternatives Assessment Dashboard for chemicals involved in several different synthesis routes for dimethyl methylphosphonate. Main reactants are highlighted in gray

	Acute Mammalian Toxicity Oral	Acute Mammalian Toxicity Inhalation	Acute Mammalian Toxicity Dermal	Carcinogenicity	Genotoxicity Mutagenicity	Endocrine Disruption	Reproductive	Developmental	Neurotoxicity Repeat Exposure	Neurotoxicity Single Exposure	Systemic Toxicity Repeat Exposure	Systemic Toxicity Single Exposure	Skin Sensitization	Skin Irritation	Eye Irritation	Acute Aquatic Toxicity	Chronic Aquatic Toxicity	Persistence	Bioaccumulation
121–45-9 Trimethyl phosphite	M	N/A	M	N/A	L	L	H	H	N/A	N/A	M	N/A	N/A	H	H	L	N/A	L	L
74–88-4 Methyl iodide	H	H	M	VH	H	L	N/A	N/A	N/A	N/A	N/A	M	N/A	H	VH	VH	VH	H	L
Methyl iodide
2857–97-8	M	L	N/A	N/A	N/A	L	N/A	N/A	N/A	N/A	N/A	N/A	N/A	VH	VH	N/A	N/A	N/A	N/A
Bromotrimethylsilane
993–13-5 methyl phosphonic acid	M	N/A	N/A	N/A	L	L	N/A	H	N/A	N/A	N/A	N/A	N/A	N/A	N/A	M	N/A	N/A	L
67–56-1	H	H	H	N/A	H	H	H	H	N/A	N/A	H	H	N/A	N/A	H	L	N/A	H	L
Methanol
676–97-1 methyl phosphonic dichloride	VH	VH	N/A	N/A	L	L	N/A	H	N/A	N/A	N/A	N/A	N/A	N/A	N/A	L	N/A	N/A	L
67–56-1	H	H	H	N/A	H	H	H	H	N/A	N/A	H	H	N/A	N/A	H	L	N/A	H	L
Methanol

The sustainable chemistry tool presented in this paper provides information on the chemical synthesis route, including feedstocks, reaction conditions, and by-products of the reaction. This information can be used not only to evaluate the candidate synthesis routes based on sustainability metrics associated with the reaction such as AE and characterize risks associated with each chemical, but also to aggregate life cycle inventory information for the chemical lineage of the materials involved in the synthesis route. Additionally, the information contained within the functional groups and reaction databases can be used to search internet databases, including the chemical literature, and form the basis for a machine learning chemical synthesis framework.

The current work is a proof-of-concept demonstration of the algorithm that applies it to search a chemical reaction database and identify relevant literature regarding reaction schemes that may be used to form the desired functional group present in a compound of interest. The current state of the database provides a minimal collection of name reactions pertinent to phosphorous compounds, with a subset of that literature focused on phosphonate ester generating reactions. However, even with an initially limited database, the flexibility of the system is demonstrated by dividing the general Michaelis-Arbuzov reaction into (3) different, more-specific reactions, based upon minor differences in reaction conditions. By segregating the reactions in this manner, the application was able to represent the diverse range of conditions the Michaelis-Arbuzov reaction can be performed under.

In the example provided, there are two variations of the Michaelis-Arbuzov reaction, along with an alkylphosphonic Acid esterification approach, and a novel solid phase approach for the synthesis of dimethyl methylphosphonate. Four references are provided for the Michaelis-Arbuzov reaction, which employ small variations in the reaction conditions. Similarly, for the esterification reaction, three different reaction conditions with references are provided, and a single reference for the solid phase reaction was provided. One challenge in expanding the literature database is defining a means to differentiate between the various similar chemical reactions within a family of name reactions to provide insight into the impact of the reaction conditions on sustainability of the molecular synthesis.

It is planned to expand the database with additional synthetic literature pertinent to the synthesis of various classes of organic compounds, including the reaction conditions and the reagents used to synthesize these compounds. The database will also provide literature references for each class of compounds which will help in identifying a sustainable chemical pathway for the synthesis of that chemical. Future work will include expanding the database by linking it with existing online literature databases and providing a means for researchers to add pertinent information about the sustainability of their approach. In this way, the database can expand, improving both the scope and applicability of the information provided.

Synopsis:

The methodology allows users to evaluate the sustainability of a compound based upon the reactions that can be used to manufacture the chemical.