Addressing Hazardous Implications of Additive Manufacturing: Complementing Life Cycle Assessment with a Framework for Evaluating Direct Human Health and Environmental Impacts

Additive manufacturing (AM) is transforming manufacturing technology and the distribution of production capital. As the use of three‐dimensional printers begins to extend into homes, schools, and factories, the industry is not well equipped to address the potential for deleterious environmental and health impacts. Proactive assessment tools are needed so that materials developers and designers, printer operators, and print end users can create and choose the most appropriate and safe materials and AM processes based on their use cases. Current life cycle assessments (LCAs) do not provide sufficient information to support materials decisions based on concerns about hazard exposure. To address this shortcoming, we developed a framework that complements LCA with hazard and green design metrics derived from analyzing human health and environmental impacts in the later stages of the AM life cycle. We then identified suitable existing methodologies for evaluation across these stages and synthesized the methodologies into higher‐level metrics for comparative analysis of materials. To illustrate the benefits of this framework, we compared two common AM materials: Autodesk Standard Clear Prototyping Resin (PR48), an open‐source formulation used in photopolymerization processing AM, and bio–polylactic acid, a ubiquitous, biosourced polymer used in an extrusion‐based AM system called fused filament fabrication.


Introduction
Additive manufacturing (AM) has the potential to reduce costs, waste, and transport associated with supply-chain activity, while increasing design complexity and customization of products, making it increasingly popular with industry and consumers. Global life cycle assessment (LCA) methodologies colleagues, its share of the ecological profile can be inadequate to address the potential magnitude of a material's impact on AM technology users and the environment (Luo et al. 1999).
The accessibility of AM technology makes industrial-level manufacturing hazard exposure more commonplace as printers scaled for home or school use become more widely adopted and a new distributed waste stream of materials is introduced, of which the reusability, biodegradability, and ecological harm is not yet well understood. Growing public concern about the hazards of materials, increasing regulatory pressure, and a growing sustainability commitment by businesses also motivate better, safer AM materials (Tickner et al. 2015). Insights into a material's hazard potential enable material developers to devise safer, more-sustainable materials while meeting performance criteria and competing on the basis of these characteristics.
Understandably, the safety and sustainability of AM materials should be considered within the broader impacts of current chemical manufacturing pathways used to produce AM and non-AM materials alike. Addressing current chemical manufacturing pathways in addition to new exposure pathways introduced by AM would provide a more comprehensive approach to chemicals assessment. However, given the enormity of such a comprehensive undertaking, the scope of this assessment will primarily be limited to the AM-related evaluation methodologies we have identified that cover material hazard, green chemistry, and ecological end of life (EoL). These methodologies will supplement LCA and methodologies evaluating exposure pathways common to AM and non-AM materials.

Identifying Life Cycle Stages for a Tailored, Informed Approach
We started by identifying life cycle stages of a material used in AM: how a material is created, used, and then dealt with at its EoL. The impact of each life cycle stage is determined by exposure pathways and informed by stakeholder requirements or perspectives. 1 Any material evaluation should also take into account the setting in which printing equipment and resulting products will be used given that this dictates the implications of material hazard. The goal in our approach is to support raw material manufacturers, material developers, product designers, printer operators, and users of printed products in making informed choices. It is not our goal to revisit LCA methodologies. Instead, we focus on materials considerations that have thus far been underdeveloped: hazards of exposure to materials during and after the printing process.
The AM material life cycle starts with extraction of raw materials from biological, inorganic, or petrochemical sources, which are processed through many chemical transformations and formulated into a feedstock for a three-dimensional (3D) printer. The printer operator uses this feedstock to produce a product that is ready for use as is or following additional postprint processing. The printed object can then be used for its intended application. Depending on the application, the 3D printed part may or may not be handled often by people (e.g., toys vs. car parts). Ultimately, the final printed object must be disposed of or recycled along with excess materials from the printing process. We group the AM material life cycle into six main stages shown in figure 1.
A full materials assessment takes into account an LCA that covers the significant impacts associated with energy output and emissions in stages 1 to 6 (Faludi et al. 2015;Diaz et al. 2012;Telenko and Seepersad 2012;Narodoslawsky 2015;Gebler et al. 2014). LCAs are important for any organization to make internal improvements based on specific needs, but generalizing this process for vastly different AM technologies is a complex task. We indicate important considerations and refer to current methodologies for evaluating the impact of how an AM material is manufactured (stages 1 and 2), but focus our comparative assessment on stages 3 to 6 because these are the stages where end users, operators of 3D printers, and the ecological sphere are most at risk of exposure to potential chemical hazards and where methodologies for assessment are novel and less developed. We also present higher-level metrics for stages 3 to 6, indicating data that stakeholders require for evaluating those metrics, followed by application of the assessment framework to compare two materials from commonly utilized AM technologies: an extrusionbased system called fused filament fabrication (FFF) and a photopolymerization process, digital light projection (DLP).

Establishing Safety and Sustainability Metrics and Criteria
We developed a three-tier scale to compare relative safety and sustainability data across stages 3 to 6 of the life cycle. A qualitative description of these levels is summarized in table 1. Level 2 describes the best-case scenario in which sustainability and safety are maximized within reasonable constraints of stakeholders. Level 0 describes the worst-case scenario for common AM technologies. Level 1 is in between, but with obvious improvement over Level 0. The thresholds between levels are dictated by the reasonable range of effects as well as a reflection of the considerations and interests of stakeholders and respective human health and environmental impacts. Both the number of levels and the thresholds for each level may be refined with a larger data set.
Comparing multiple AM materials by life cycle stage makes it easy to identify differences in impacts at each stage. Thus, parsing out the different stages in the AM life cycle provides a more precise understanding of the hazards encountered and a more targeted approach to improving safety and sustainability.

Overall Life Cycle Assessment
In order to effectively supplement AM LCAs, we consider their value and shortcomings. Previous investigations of the  environmental effects of AM have explored claims that the technology could reduce the carbon footprint of products through the reduction of transport and supply-chain activities (Faludi 2016;Kreiger and Pearce 2013). Luo and colleagues (1999), Gonzalez (2007), and Faludi and colleagues (2015) utilize such metrics to evaluate the impacts of AM materials sourcing. Narodoslawsky (2015) and Tabone and colleagues (2010) also present a valuable analysis on how to use LCAs to determine the environmental impact of sourcing chemicals in general. Most of these studies focus primarily on energy use or other factors (Kreiger and Pearce 2013;Behrendt et al. 2012). Efforts by Gebler and colleagues (2014) and Faludi and colleagues (2015) have addressed the environmental impacts of AM more exhaustively. In Faludi and colleagues (2015), this involved tracking all major impact types by comparing AM to machining of hollowed-out thermoplastic parts, monitored from cradle to grave (Faludi et al. 2015). 2 The main conclusion of the study was that, in contradiction to earlier hypotheses, the ecological impacts of transportation, disposal, and material concerns paled in comparison to printer energy use for most printers. This would suggest that the highest priority for increasing the sustainability of AM would be to make 3D printers more efficient in their use of electricity. But, as printers become more efficient, the contribution of materials choices to overall ecological impacts and human health becomes much more significant, making analysis and comparison of impacts related to specific materials more important. We recommend performing LCA in accord with the International Organization for Standardization (ISO) standards-ISO 14040:2006 and ISO 14044:2006-to do such analyses.

Raw Materials Sourcing, Manufacturing, and Formulation (Stages 1 and 2)
AM raw materials are extracted from the earth, derived from bio-based sources, or reused from existing materials (Xiao et al. 2012). Bio-based and recycled material sources often appear to be the more attractive option because of their adherence to green design principles, but a detailed LCA reveals that some biologically based polymers may present a greater environmental impact (Kreiger and Pearce 2013;Lithner et al. 2011). Dramatically different environmental profiles emerge based on farming location or the process required to transform the biological material into AM printing material feedstock (Kreiger and Pearce 2013). No single petroleum source is equal eitherthe type, location, and extraction method of fossil fuel feedstocks all contribute to the resulting emissions profile (Skone 2011).
Green chemistry principles provide a different perspective that chemical manufacturers and the supply chain are beginning to implement (Alfonsi et al. 2008;Dunn 2012;Giraud et al. 2014). Green chemistry is defined by the U.S. Environmental Protection Agency (US EPA) as "the design of chemical products that reduce or eliminate the use of hazardous substances" (Dunn 2012(Dunn , 1452Giraud et al. 2014). Green chemistry principles most relevant to industrial manufacturing of chemicals are (1) reduce waste, (2) prevent hazardous chemical exposure, (3) choose more-efficient reactions, and (4) use less energyintensive processes.
Even though hazard exposure to chemicals used is less of a concern during the sourcing and manufacturing stages than for later life cycle stages attributable to engineering controls and personal protective equipment used in industrial settings, toxicity and environmental traits are still important in case of accidents or leaks, which can affect the ecosystems, communities, and workers where these materials are produced or sourced (Lent 2003). For example, data from the European Union (EU) found that improving the safety of workplace chemicals could prevent up to 40,000 cases of asthma annually, an equal number of dermatitis cases, and 10,000 cases of chronic obstructive pulmonary disease each year (Pickvance et al. 2005). To address these considerations, an appropriate methodology for evaluating sourcing and manufacturing of materials should include metrics from green design principles and hazard evaluation methodologies to supplement an overall cradle-to-cradle LCA. Green design metrics include those that evaluate waste, reaction yield, material recovery parameter, reaction mass efficiency, atom efficiency, and stoichiometry. Several methodologies combine these factors into an overall rating. For example, in the GlaxoSmithKline (GSK) reagent selection guide's "Clean Chemistry" score, the green design factors mentioned above are weighted equally to give an overall "Clean Chemistry" rating (Henderson et al. 2011). The EATOS (Environmental Assessment Tool for Organic Synthesis) has also been commonly utilized. 3 Comprehensively evaluating hazards in sourcing and manufacturing materials is precluded by the large number of chemicals that are used in the manufacturing of one material; extensive chemical assessments of all chemicals that make up a material is not always feasible (Gauthier et al. 2015). Instead, as a first-pass approach, chemicals can be grouped into two categories: reagents and solvents, and evaluated more generally (Prat et al. 2014). Henderson and colleagues (2011) devised a methodology to compare the hazard potential of solvents used in chemical manufacturing. A solvent is evaluated based on several criteria: waste, environmental impact, health, flammability, reactivity, and life cycle score and rated on a three-tiered scale (Henderson et al. 2011). Assessing the potential hazards of reagent chemicals is more specific. Most chemicals used in manufacturing have been categorized according to Hstatements, a standardized classification system to indicate the nature of hazard and route of exposure. McElroy and colleagues (2015) categorizes chemicals in manufacturing according to these H-statements toward an overall score of high, medium, or low.
In summary, when scoring the materials sourcing, manufacturing, and formulation stages in the AM material life cycle, LCA metrics, green chemistry principles, and hazard should all be taken into account to give a comprehensive view of the sustainability of material choice.

Printing Process (Stage 3)
The most conspicuous AM materials hazards are usually presented to the operator and those in the surroundings of an AM technology during the printing process. Exposure pathways and hazards vary somewhat depending on the type of process. Gibson and colleagues (2015) separate AM technology into seven different processes: (1) photopolymerization processes; (2) powder bed fusion processes; (3) extrusion-based systems; (4) printing processes; (5) sheet lamination processes; (6) beam deposition processes; and (7) direct write technologies. All of these methodologies present some form of hazard to the operator of the printer. These hazards include contact (or dermal) and environmental hazards, physical hazards (potential for explosion or fire), particulate emissions (including ultrafine particulate emissions or UfP emissions), and volatile organic compound (VoC) emissions (Azimi et al. 2016). Postprocessing techniques are highly variable-even within a single type of technology-and they contribute to additional potential hazard exposure (Gibson et al. 2015). Although all of these hazards may be present for each of these AM technologies, the potential for exposure varies among the seven technologies. For this assessment, we will focus on the first three technology types given that most research conducted on the hazard of AM has centered around these technologies.
Extrusion-based systems, such as FFF, and powder bed fusion processes, such as selective-laser sintering (SLS), can generate chemicals from thermal decomposition as well as particulate emissions. For example, recently, Kim and colleagues (2015) found that FFF 3D printers, regardless of the material used, emitted nano-size particles at high concentrations, several aldehydes (including carcinogenic formaldehyde), phthalates, and VoCs such as toluene and ethylbenzene. In another example, Graff and colleagues (2017) recently determined that there was significant operator exposure to metal particles larger than 300 nanometers (>10 8 particles per cubic meter over 10-second cycles) when handling metal powders for SLS printing. Whereas welding processes, laser printers, and toasters have been known to produce similar concentrations of particulates, the toxicological profile of nanoparticles produced from AM plastics by FFF as well as certain metals like cobalt for SLS could be more harmful (Oberdörster et al. 2005;Peixe et al. 2015;Andujar et al. 2014). FFF and SLS can also present potential physical hazards, including explosion and fire hazards.
Lower-temperature photopolymerization processes, like stereolithography (SLA) and DLP, typically utilize liquids that are not easily aerosolized and seldom contain components with significant vapor pressures (Carroy et al. 1997). Yet, because these materials are very reactive, they tend to be more toxic to humans and aquatic life and their bioavailability tends to make them worse environmental toxins.
Postprocessing procedures are often used after printing to make a printed object ready for use (Gibson et al. 2015). SLA and DLP prints require washes with organic solvents, such as alcohols or propylene carbonate, to remove residual resin (monomers) from the printed parts. Solvent baths are sometimes used in FFF and SLS for removing supports or improving surface quality. Other postprocessing processes include sand blasting, sealing parts, electrochemical polishing, bonding, and gluing (Gibson et al. 2015). A summary of the potential hazards for three common AM processes appears in table 2.
A comprehensive assessment addresses each hazard concern by assigning a rating of a material's contact and environmental hazards, potential physical hazards, and potential for exposure to VoCs and UfP emissions, respectively, based on the particulars of the process used. In each category, a material may be scored between 1 and 4. A score of 4 is designated to compounds with low concern, whereas 1 is designated to compounds with high concern; most hazard-rating methodologies utilize a fourtiered scale. See tables S6 to S12 in the supporting information available on the Journal's website to understand how each score is attributed for each criterion. For example, SLA and DLP involve little risk of exposure from physical hazards and UfP emissions, whereas SLS and FFF present greater potential for physical hazard and exposure to UfP emissions. Operator risk can be reduced by engineering process controls and personal protective equipment or eliminated by using less-hazardous materials.
We propose the following approach to determine the material contact and environmental hazards in the printing process. Start by screening for known hazardous chemicals found on authoritative and regulatory lists of restricted chemicals in databases such as Healthy Building Network's Pharos Chemical and Material Library (Lent 2014), SINLIST, 4 and SUBSPORT. 5 See the Supporting Information on the Web for more information on these searchable databases. The next step is to gather data from safety data sheets, which are created to provide workers and emergency personnel with procedures for handling a substance in a safe manner. Once data are compiled, several methodologies are available to rate materials for contact and environmental hazards posed during printing. One example that enables rating each chemical in a materials formulation is GreenScreen-a comprehensive methodology to understand the toxicity profile of a material (Heine and Franjevic 2013). A GreenScreen rating involves grouping hazard, organizing toxicity data into those groupings, and then evaluating a chemical based on its combination of toxicity ratings. Chemicals are then ranked from Benchmark 1 (a chemical of very high concern) to Benchmark 4 (a chemical of low concern). Less-rigorous processes, such as GreenScreen's ListTranslator and QCat, 6 will nevertheless provide a first-pass analysis to identifying hazard "hotspots." If no hotspots are identified, a full GreenScreen assessment, which requires more-comprehensive data mining or computational toxicology assessments to fill in data gaps, provides full certainty of a chemical's safety.
Unfortunately, none of these approaches address how to evaluate mixtures of chemicals. This is problematic because most AM materials are comprised of several chemicals. Two key considerations must be taken into account to address chemical mixtures: (1) antagonistic or synergistic interactions between chemicals in the mixture and (2) dose-response relationship of chemicals in the mixture (e.g., some chemicals exhibit similar toxicity in low concentrations of the material/mixture whereas others register undetectable toxic effects in small concentrations). For simplicity, we do not take into account chemical interactions, especially because at lower dose levels these interactions are toxicologically insignificant. We propose a method for evaluating mixtures of chemicals while accounting for doseresponse relationships by factoring in the percent of each chemical in the material (Autrup et al. 2012). If a material contains any GreenScreen Benchmark 1 chemicals >100 parts per million (ppm), then the material is automatically given a Contact and Environmental Hazard score of 1. This is because Benchmark 1 chemicals often do not have linear dose-response relationships; these chemicals have similar effects in very low concentrations as they do in higher concentrations. If a Benchmark 2 or 3 chemical comprises <10%, the material may be given a score one rating higher. If a Benchmark 2 or 3 comprises >10% of the material, then the material is rated by the lowest Benchmark chemical that comprises >10% of the material. This methodology is summarized in table S6 in the supporting information on the Web. Postprocessing evaluation takes into account the amount of waste produced, energy used, and hazard of auxiliary chemicals used in postprocessing. Waste of a material may be addressed by the concept of "E Factor" (Sheldon 1997). The E Factor is based off the process mass intensity shown by equations (1) and (2):

PMI = Mass of all materials used to make the product Mass of product
Energy intensiveness of postprocessing is outside the scope of this framework, but we encourage future studies to compare the energy output of various postprocessing techniques. The hazard of postprocessing techniques can be determined from Green-Screen Benchmarks (similar to the contact and environmental hazard assessment outlined above). The Post-Processing score may be determined from these three considerations (see table  S8 in the supporting information on the Web) and given a score ranging from 1 to 4.
In summary, for the Printing Process stage, materials are rated on their physical hazard, contact and environmental hazard, the potential for exposure from UfP emissions, the potential for exposure from VoCs, and their postprocessing score. Score thresholds for each criterion are dependent on the AM technology used, and materials are given an overall score between ranging from a Level 0 to Level 2. These score thresholds are defined in Tables S11 and S12 in the supporting information on the Web. A Level 2 material is one that completely mitigates most of these concerns, making it a consumer-friendly material. A Level 1 presents some concerns to an AM technology operator, and some engineering controls or personal protective equipment may be required. Last, a Level 0 material is one that cannot be used with an AM technology without significant engineering controls. With additional examples of materials evaluated, these thresholds may change to more accurately reflect the true range of material toxicity during the Printing Process stage.

Print Use (Stage 5)
While AM for prototyping remains useful, AM is increasingly used to produce marketable products. At the same time, hazards of this stage of the AM life cycle appear to be the least understood; risk depends on how the resulting 3D prints are used and in what setting (household, industrial, etc.). In the Print Use stage, we decided that the most appropriate categories to consider based on common uses of AM materials would address applications such as wearables, toys, household products, and industrial parts.
Analyzing the material hazard in the Print Use stage can be difficult using standard toxicology review methodologies, such as GreenScreen, given that the chemical composition of the final product is often different than the material used in the printing process, attributable to chemical changes or contamination from the manufacturing process (Heine and Franjevic 2013). Because the chemical constituents are not all known, chemical composition testing would be required as well as another round of toxicology assessments of the new individual chemical constituents. In a recent example, a chemical screen of a sample of a pizza box revealed many unidentifiable compounds-this demonstrates the uncertainty in content of even industrially produced products (Bengtström et al. 2016). With this complexity in mind, we agree with John Warner, a strong advocate for product chemical safety, who recently laid out an argument for standard testing of finished products (Warner and Ludwig 2016). If the content and safety of everyday products is a concern, 3D printed parts may pose more concerns because of the distributed nature of use and resulting lack of supply-chain knowledge regarding how materials and the resulting finished printed parts will be used. A finished product testing schema that rates a product based on its performance remedies these concerns by allowing a printer operator to identify and control for product chemical safety issues without having to convince suppliers to reveal trade secrets.
Because there is no single standard to understand or estimate the potential hazard in a product that includes materials, we looked at a set of authoritative lists and regulations to derive a list of qualitative and quantitative metrics for creation of this portion of the assessment framework. The U.S. Food and Drug Administration food contact regulations, the emerging guidance from Underwriters Laboratory (UL) for wearable tech industry and Nestlé Standards on Materials in Contact with Food (GI-80.008) include relevant tests for certain applications. ISO 10993 7 describes the procedures necessary to test medical devices, which are often in intimate and prolonged bodily contact. Whereas medical device safety standards usually exceed what is needed for most AM applications, the test methods in ISO 10993-1 4 for external communicating devices (articles that are used in direct contact with undamaged skin) provide useful guidance for more ordinary, less-regulated uses of AM such as wearables. We consolidated this information into a simple set of indicators for toxicity that incorporate currently available testing methods already in use. These indicators are listed in table 3 and are based on an avoidance of restricted chemicals from authoritative lists as well as outcome-based toxicity considerations such as cytotoxicity, irritation, and sensitization. In our framework, Level 2 prints are deemed safe by these indicators for prolonged skin contact (greater than 24 hours, i.e., a watch, or a bracelet) and limited short-term oral contact (i.e., a spoon), whereas Level 1 materials are safe to use with intermittent skin contact (greater than 1 hour, but less than 24 hours), and Level 0 materials are presumed unsafe for direct skin or oral contact unless direct chemical testing confirms safety (for less than 1 hour). This first step involves doing a full chemical-characterization of a print and verifying whether any identified chemicals (in concentrations greater than 100 ppm) are restricted in accord with various authoritative lists or regulations or classified as substances of very high concern (SVHCs) as dictated by the Registration, Evaluation, Authorization and Restriction of Chemicals (REACH) regulations (Lent 2014). If any of these chemicals are found above these concentrations, then the material is considered a Level 0 material for toxicity. The next step is testing for specific toxicity considerations, starting with cytotoxicity (Li et al. 2015). The ISO 10993 methods include cytotoxicity tests that rapidly and costeffectively screen materials (Mallikarjuna 2014). They are well established for dental materials, which are compositionally very similar to SLA materials (Nedeljka et al. 2013;Goiato et al. 2015). However, cytotoxicity tests are not without problems. For one, ISO 10993-5 does not prescribe any one specific test method or protocol (although examples are given in ANNEXs A to C). 8 Because it is unclear from ISO 10993-5 which set of cytotoxicity tests are most appropriate, we leave the choice of test methods open, but also point to methods for evaluating dental materials that leach similar classes of materials as SLA resins (Ivković et al. 2013). To obtain a Level 2 classification for materials, irritation and sensitization tests are also required and must pass those tests as described by methods in ISO 10993-10. Last, if any chemical in the material is found to be a dermal toxicant, according to testing methodologies outlined in ISO-10993, it is characterized as a Level 0 material; only oral toxicity results in a Level 1 designation. This discrepancy is attributed to the fact that, generally, oral exposure pathways are less common than dermal exposure pathways.
Other important toxicity considerations for AM printed materials concern toxic metals and other known toxicants. Regulations restricting the presence of heavy metals and phthalates, such as The Consumer Product Safety Improvement Act, 9 the Toxics in Packaging House, 10 U.S. Public Law 104-42, 11 and similar European laws, provide thresholds for specific heavy metals: cadmium, chromium, lead, and mercury, based on pre-existing health studies and research. Taking these legal requirements into account, our framework recommends that prolonged contact (Level 2) requires concentrations of these metals to be under 1,000 ppm. UL's guidelines for wearable electronics also suggest limited nickel leakage to 0.5 micrograms per square centimeter per week (µg/cm 2 /week) (UL "Outline of Investigation for Sustainability . . . " 2014). 12 In addition to metals, polybrominated diphenylethers (PBDEs), polybrominated biphenyls, dibutyltin, dioctyltin, pentachlorophenol, and azo-dyes should all be limited according to concentrations indicated in table 3.  Note: PLA = polylactic acid; AM = additive material; UfP = ultrafine particulate; VoC = volatile organic compound; n/a = not applicable.

Printing Process Waste Disposal and Print Disposal (Stages 4 and 6)
The principal concerns related to disposal of AM outputs (prints, wasted feedstock, and other process by-products) are not immediate exposure to hazard, but the long-term environmental impact. Even though hazardous waste must eventually be treated according to local environmental regulations, these treatments are often costly, energy intensive, and do not completely eliminate the risk of environmental contamination (e.g., groundwater leakage, air pollution) (Wilson and Schwarzman 2009). AM enables access to distributed manufacturing models in developing countries, but proper hazardous waste disposal options also are not widely available in many of these countries, especially when waste is generated in smaller quantities at dispersed locations (Mmereki et al. 2016).
Although the material configuration during and after the printing process may be quite different, the criteria used to evaluate the safety and sustainability are very similar. In our framework, the only criterion that differs is whether waste is produced during the print processing; if no waste is produced, then a material will automatically achieve a Level 2 rating for the Printing Process Waste Disposal stage. Otherwise, at the EoL stage, unused printing process materials or prints should be either reusable for AM, biodegradable 17 (through composting or other means), or recyclable 18 to meet our criteria for a Level 2 rating. AM is becoming more multimaterial, which makes single-stream recycling more difficult, if not impossible . This suggests enhancing biodegradability as the most viable option. If a material cannot biodegrade or be reused in some fashion, but is not considered hazardous waste according to US EPA 19 or California Department of Toxic Substances Control (DTSC) 20 regulations, it can at least be safely landfilled for disposal and will be considered a Level 1 material. Level 0 materials are considered hazardous waste, according to those same regulations, or material that degrades into hazardous decomposition products. Figure 2 details this decision-making process in a flow chart. A compilation of research on certain material degradation pathways may provide guidance for understanding decomposition products. For example, common thermoplastics, like acrylonitrile-butadienestyrene and polystyrene, can decompose into cancer-causing styrene in warm ocean gyros (Tullo 2016). Other plastics, such as polypropylene, absorb toxic compounds already in the environment, such as PBDEs and polychlorinated biphenyls (PCBs), as they break down into microplastics (Rochman et al. 2013).

Results: An Illustration of This Framework
In order to demonstrate how one can use this framework in a real-life application, we analyzed two different 3D printing materials, bio-PLA (polylactic acid) (from sugar cane) for extrusion-based FFF printing and Autodesk's PR48 for a photopolymerization-based printing. We applied our methodology and analyzed the materials against each metric to compare the results. For the purposes of this article, we will explain the analysis on the two materials for one stage of the process, the Printing Process stage (3), and then give information on the results of an overall evaluation for stages 3 to 6. Tables S14 and S16 in the supporting information on the Web describe these results in greater detail.

Printing Process Stage Example: Polylactic Acid and PR48
Before assigning each a material a Level (0, 1, 2), scores for each hazard category are determined (table 4). When compared to the metrics in this framework stage, bio-PLA achieves a contact and environmental score of 3 (out of 4) because residual contaminants (<1%) that are often in this kind of material may have significant reproductive toxicity hazard (see sections B and D in the supporting information on the Web). A score of 2 (out of 4) for the postprocessing category is based on FFF techniques that are commonly used that involve large energy intensiveness and hazardous chemicals. UfPs and VoC scores of 4 and 3, respectively, are derived from recent studies that show comparatively low levels of these compared to other thermoplastics. Because every score is greater than or equal to 2, the result is a Level 2 designation for the Printing Process stage (refer to table S11 in the supporting information on the Web for assigning overall scores to a designated level).
PR48 is given a score of 1 (out of 4) for the contact and environmental hazard criterion attributed to its significant aquatic toxicity and sensitization potential. It achieves a 3 rating for postprocessing because a significant amount of waste is generated from the supports needed for the printing process and a large amount of organic solvent (isopropanol) required for postprint washing. The VoC emissions profile is insignificant for printing in close quarters, resulting in the safe rating of 4. The very low contact and environmental hazard criterion score is the main reason for PR48's Level 1 designation for the Printing Process Stage (refer to table S10 in the supporting information on the Web for assigning overall scores to a designated level). Individual criterion scores and the overall score are summarized in table 4.

Comparing Polylactic Acid and PR48 at Different Stages
To illustrate how to compare different AM materials across multiple stages, we demonstrate a summary of a comparison of PLA and PR48 materials in stages 3 to 6, from Printing Process to Print Waste Disposal. The attributed levels for each stage will be combined by a weighted average, resulting in an overall score ranging between 0 and 2. This weighting can be modulated, though, to highlight the importance of one stage over another in accord with the decision maker's or stakeholder's needs. For example, if one is considering a children's toy produced from AM, one may want to weight safety metrics in the Print Use stage as more important than other stages when selecting between materials. The varied depth of analysis in this assessment approach permits a direct comparison of materials across each stage of the AM life cycle. From the summary in table 5, it is apparent that the overall low score of PR48 relative to PLA is attributed to lower scores across the board, but particularly in the Print Use and Printing Process Waste Disposal stages. This summary analysis suggests prioritizing those hazard categories and components to make the biggest difference in designing materials that are safer and more sustainable. A more in-depth analysis would involve analyzing the Printing Process stage, for example, and discovering that PR48 has a lower score than PLA because it rates higher in hazard for certain toxicity categories (e.g., like sensitization). It is clear that the ratings differences between PLA and PR48 are quite stark; differences in ratings will likely be smaller between AM materials from the same AM process.

Conclusion
Our goal for creating and publishing this assessment framework is to facilitate transparency and cooperation in AM materials development and use, while minimizing their hazards and environmental impacts. We believe simpler metrics and shared tools for communicating and comparing options will improve AM materials development and use. In this article, we have described a framework to incorporate potential hazards and environmental impacts that complement an LCA of an AM material, particularly from printing onward. In developing this framework, we show that a full assessment of all the impacts of AM material choices requires an LCA along with methodologies that take into account direct hazards to people throughout the life cycle. Further, while many gaps exist in the data sets examined, there are no areas where analytical methods are absent. LCAs provide insight into the origin of the printing materials and potential global impacts; GreenScreen and other hazard methodologies provide identification and evaluation of hazard during the materials manufacturing and then printing process; ISO 10993 provides guidance for using prints; and existing disposal regulations are appropriate for the end stages. Finally, from a preliminary analysis of two materials, it is apparent that handling uncertainty is a challenge, but that the multilevel analysis does indeed allow one to pinpoint opportunities to improve a process or material.
In sharing this framework, we hope to create a path forward for safer, more-sustainable materials and allow others to build on our work with additional data, feedback, and analysis. However, the main function will remain intact: a multilayered transparent approach that organizes access to information and corroborated data for varying depths of analysis. We expect this framework will be used differently by those seeking to deploy AM in novel ways, those seeking to design better materials and processes, and end users of 3D printed products. Much of our analysis was informed by extrusion-based technologies, photopolymerization processes, and powder bed fusion processes. The world of additive manufacturing is much broader than these technologies and changing rapidly. As such, the tools for understanding it will need to continuously evolve as well.