Proceedings of a Workshop—in Brief

January 29–30, 2025

Mid-Scale Manufacturing and Characterization Capacity for Department of Defense Critical Materials Supply Challenges, Part 1

On January 29–30, 2025, the National Academies of Sciences, Engineering, and Medicine’s Defense Materials, Manufacturing, and Its Infrastructure Standing Committee hosted the first part of a two-part workshop sponsored by the Department of Defense (DoD). This workshop was aimed to examine U.S. manufacturing and characterization capacity for mid-scale production. Strategic and critical materials are vital to U.S. national defense and economic prosperity, enabling the United States to develop and sustain emerging technologies and improve its warfighting capability. DoD defines strategic and critical materials as those that support military and essential civilian industry and are not found or produced in the United States in quantities to meet its needs. Hence, accelerating, scaling up, and transitioning technologies to produce or replace critical materials, some novel, are essential to mitigating DoD supply challenges. Mid-scale manufacturing and characterization capacity in combination with modeling and simulation are expected to play a key role in this effort. Key workshop topics included the availability, access, and economic sustainability of both mid-scale manufacturing facilities and experimental facilities that provide extreme environment characterization, with particular attention toward defense-specific applications. The workshop events¹ also explored state-of-the-art approaches used to evaluate data, models, and simulations for scale-up. Participants’ presentations and discussions from the first workshop, which focused on manufacturing opportunities and drivers, are summarized below.

A STRATEGIC DECISION-MAKING FRAMEWORK

Angus Kingon, Brown University, described mid-scale manufacturing as a key step in the commercial development process, focusing on manufacturing readiness levels (MRLs) 7–8 and technology readiness levels (TRLs) 6–9. In the context of DoD specifically, mid-scale manufacturing refers to the manufacturing of strategic quantities (i.e., relatively low volumes in final production) of military components or systems such as specialty alloys and energetic chemicals.

Kingon explained that the decision to commercialize is often driven by economic opportunity. Many barriers exist to commercial viability, however, with both market and technical risks, and understanding economic opportunity begins by assessing customer needs. Another driver for commercialization is national strategic need, which focuses primarily on technical risk and scale-up potential.

and integrated device manufacturers (IDMs), making GaN easier and cheaper to purchase in the future. He reiterated that Qromis would like to participate in establishing manufacturing for this critical technology domestically—and niche applications could be useful for DoD—but previous attempts to do so were unsuccessful. Mark Johnson, Clemson University, indicated that the United States had a strategy for GaN in 2000–2005, but the Defense Advanced Research Projects Agency (DARPA) decided to focus primarily on SiC. He noted that this example demonstrates the limits of public-sector funding and the potential ramifications of non-market-driven decision making. Richard Vaia, Air Force Research Laboratory, remarked that one government agency’s decision should not disrupt the entire ecosystem; for example, the Department of Energy has continued to invest in GaN. Kingon asked if a clear path forward exists for U.S. companies to invest in innovative GaN devices identified by universities. Basceri said that very limited GaN activity and GaN device players exist in the United States (e.g., EPC, Navitas, and Vertical Horizons as fabless companies; Power Integrations as an IDM; and Global Foundries as a foundry without a qualified GaN offering), and all of them rely on either non-U.S. manufacturers or on limited-performance substrates (Si, sapphire) as well as non-CMOScompatible, expensive platforms. Overall, he continued, many opportunities exist by leveraging the current semiconductor manufacturing platforms in the United States.

Vaia asked what the federal government could have done to accelerate the P/C curve for GaN. Basceri noted that DoD is a potential and important customer for GaN; as such, a DoD mandate, as one of the legs of the dual-use approach, will likely further ignite GaN activity in the United States. There are Tier 1 power semiconductor manufacturers in the United States, such as Texas Instruments and onsemi, that might be interested in DoD’s co-investment proposal into this critical supply chain. Another participant wondered how supply chain factors into economic viability for GaN, given recent material and import restrictions. Basceri emphasized that such restrictions represent a global issue, not just one facing the United States. He added that although direct impacts may include increased cost, and those effects might be temporary, these might be not as critical now because GaN production volumes are not as high as Si-based solutions yet.

Rudy Wojtecki, Applied Materials, highlighted the competitive nature of the field; he inquired about the lead time before others might try to copy QST^® and whether a strategy for enforcing (IP) exists. Basceri indicated that the lead time is a minimum of 2–3 years, and Qromis’s IP is very extensive, but legal action often ensues when people attempt to copy the technology, as in all other IP-protected innovations and technologies. Basceri also added that trademarks are key for recognition and for representing reliable, true-origin products. James Conley, Northwestern University, asked whether Qromis’s technology is licensable to chip design companies, like Synopsys. Basceri replied that Qromis is open to enabling further QST^® and GaN-on-QST^® adoption in many different ways, including providing additional licenses; however, there is a very fine balance between being a pure licensing company or being a product-plus-licensing company. Currently, Qromis has chosen the latter, considering the stage of GaN evolution and Qromis’s growth strategy. With that, under careful consideration, Qromis may choose to offer additional licenses to strong materials and device companies (e.g., IDMs, foundries). This offering can be an essential piece for establishing a domestic supply chain by enabling the existing U.S.-based semiconductor manufacturers with their internal QST^® substrate and GaN-on-QST^® device production under license from Qromis.

NEXTFLEX AND HYBRID ELECTRONICS

Dan Gamota, NextFlex, defined hybrid electronics as a portfolio of capabilities that use additive electronics processing and bare semiconductor die to create flexible and conformal systems, structural electronics, and additive microelectronics architectures. Hybrid electronics enable the design of novel functionality, mass customization and specialization, and higher complexity systems. In addition, hybrid electronics processes allow for on-demand and point-of-need manufacturing, enhanced process flow configurability, agility and scalability of manufacturing volumes, manufacturing convergence of semiconductors, heterogeneous integration, microelectronics processes, and democratization of distributed and networked manufacturing. Typical application areas include those that

need to be flexible, conformal, stretchable, biocompatible, lower-capital-expenditure and lower-operational-expenditure barriers to entry, lightweight, low-profile, miniaturized, cost-efficient, digital, rapidly designed, commercially launched, and sustainable. The hybrid electronics portfolio of capabilities spans semiconductor back-end-of-line assembly and packaging, heterogeneous integration, printed circuit board/assembly, and final complex system integration.

Gamota pointed to a significant increase in the number of both DoD and commercial customers requesting hybrid electronics. Commercial companies in particular want to ensure that the supply chain is robust. However, he stressed that the same rigorous engineering and investment are needed no matter the volume desired; volume simply relates to the capacity of the facility needed. He added that hybrid electronics are dual use. Whether for DoD or commercial use, the customer product functionality and performance demands are the same even when the manufacturing volume demands differ. For example, a DoD and a commercial land vehicle both require vehicle communications and data connectivity; vision systems; driver information; central compute; and control units that have low profiles, have small footprints, and are lightweight.

Gamota explained that the Manufacturing USA institutes⁴ in particular are trying to revitalize open collaboration and accelerate innovation. As an example, the capabilities within the NextFlex institute would have been accessible to employees in vertically integrated companies such as Motorola and IBM prior to 2005. However, since that time, divestitures and outsourcing led to a transition from open to closed access to capabilities, which has stifled innovation. As one of the 18 Manufacturing USA institutes, NextFlex strives to create a strong U.S. industrial base for hybrid electronics manufacturing by advancing the technology and manufacturing capabilities, establishing and growing the manufacturing industrial base, and securing the critical human capital needed for the manufacturing capacity. Gamota indicated that NextFlex has a technology hub, with 10,000 square feet of laboratory space,⁵ and serves as a consortium, with more than 200 members and partners in materials, equipment, manufacturing, and products. NextFlex also has 11 technology working groups with 5-year roadmaps that validate market needs; nationwide workforce development programs; workshops and webinars; and project calls for miniaturization, integration, ruggedization, and automation to accelerate hybrid electronics innovation.

Gamota emphasized that NextFlex focuses on addressing the “valley of death”—the time between early research and mass production—with design, manufacturing, test, and validation. He reiterated that these steps are essential and time consuming no matter the volume desired; manufacturers who will scale the products want to be confident that their launch has been thoroughly derisked. Some of the hybrid electronics–enabled systems that NextFlex has prototyped for handoff to manufacturers include solutions for security access and control, condition-based maintenance, personal physiological monitoring, and in-theater platoon situational awareness. Furthermore, he said that all NextFlex equipment sets can exchange data; having a complete digital twin improves modeling and accelerates decision making.

In closing, Gamota suggested that awareness of opportunities for hybrid electronics could be increased by (1) publishing a comprehensive report for the status of technology maturity, adoption, and key transition benefits as well as developing derisking strategies for technology, manufacturing, markets, and co-investments; (2) generating a portfolio of enabling capabilities with a baseline toolkit and advanced product research, development, and engineering; and (3) strengthening the hybrid electronics industrial base by establishing a resilient supply chain, facilitating democratization, accelerating capital expenditures innovation, and encouraging collaboration—especially to prepare for operational excellence in space and the use of autonomous maintenance platforms.

high-throughput, self-driving laboratories. Second, regarding the issue of creating features versus products, he described the Creative Destruction Lab,⁷ which helps science-based startups to scale via structured mentorship. He pointed out that when early-stage ventures do not translate into entrepreneurial success, that failure might be due to a lack of high-quality business judgment, not a lack of ideas, capital, or effort. He also described the company SRTX, which started as a materials science company that made Sheertex^® knit, a “feature,” but framed itself as a company with a “product” of “bulletproof” tights. This transition allowed the company to have a market and secure private-sector funding. Third, to address potentially long lags to commercialization, he said that early investments by the government (via contracts or prizes) could enable companies to succeed. Fourth, he reiterated that the most innovative industrial changes require coordination. He introduced the concept of “general purpose technologies,” which describes technologies that have transformative impact on the economy (e.g., electricity); however, this effect is only possible with the success of complementary innovations (e.g., the light bulb). He asserted that AI will likely fit into this category, while materials being developed today likely will fit into the category of “enabling technologies,” which could make a significant difference and lead to innovation once used in a product.

Goldfarb added that for many products, commercialization success depends on the actions of many other entities. He suggested that government and academia encourage coordination along the supply chain for certain technologies—for example, standards-setting organizations could enable a single factory to support several startups. In closing, he noted that a national security imperative exists for domestic manufacturing capabilities. If domestic production cannot compete in a free market, however, he suggested the use of subsidies, taxes, and quotas.

Discussion

Johnson posed a question about the relationship between commercialization and business models. Goldfarb described three materials science commercialization paths: (1) companies discover a material, want to create it, and want to scale and sell it; (2) companies discover both a material and an application, and want to sell directly to end users; and (3) companies serve as a platform, with a process that could be used to discover a material. The biggest market opportunity is to focus on software, he said, but IP challenges exist and a platform only emerges after the creation of two successful products.

Basceri inquired about the advantages and disadvantages of fabless companies. Goldfarb explained that fabless companies do not have to spend money building infrastructure or worry that infrastructure will become outdated. However, such companies are dependent on others for manufacturing. He highlighted the potential role for government or other buyers to support multiple entities in building factories.

Hattrick-Simpers asked how to convince people outside of academia that their investments of time and money are worthwhile. Goldfarb replied that even in the case of low-risk products, 80 percent of new products fail. Risks relate to both the demand and the scaling potential of the technology, he continued, and the private sector only wants to invest if the technology has been derisked, will lead to a use case, and will make a profit. Tom Kurfess, Georgia Institute of Technology, inquired about the relationship among a business plan, manufacturing, and scaling. Regarding investment in manufacturing facilities, Goldfarb noted that a business plan should convince a private investor that the technical risk is low and the market is substantial, and no one else has the right skills, materials, and IP for the task. He pointed out that manufacturing overseas is likely cheaper than manufacturing domestically, however, which creates a challenge for the business plan. Another cost could arise with the need for additional engineering talent to manufacture domestically, he continued, but the private sector could still be incentivized. Corbin Vara, Applied Research Associates, asked how to convince policy makers that investment in materials science is a public good. Goldfarb highlighted the need to understand the nature of the “public good.” For example, ideas, materials, discovery, and scaling capability all could be considered public goods. He suggested framing basic research as a public good for all of humanity, and input into

essential industries as a public good to preserve national interest and security for U.S. citizens.

PANEL ON SEMICONDUCTORS, MANUFACTURING, AND PHARMACEUTICALS

John Kelly, IBM, presented a successful case study from the semiconductor industry. He noted that the valley of death can be overcome via public–private partnership across the supply chain, large scale, and a sustainable business model. One public–private partnership that focuses on bridging this gap is NY CREATES (New York Center for Research, Economic Advancement, Technology, Engineering and Science), which was established in 2001 by IBM and the State of New York and is run by the State University of New York at Albany and the state’s research foundation.⁸ Multiple partners leverage shared facilities via the continually expanding Albany NanoTech Complex, working on advanced logic process technologies, memory technology neuromorphic computing, nano-bio devices, packaging technology, heterogeneous integration, integrated photonics, quantum technologies, and power electronics. Approximately $30 billion has been invested in the complex, attracting the necessary talent from around the world. Kelly explained that the entire supply chain of the chip industry essentially works together via this single location, while a satellite facility in Rochester focuses on advanced packaging. He noted that the 22 nm CMOS technology was the first to emerge from this partnership, and by 2021, the partnership was responsible for inventing and producing the first 2 nm devices, among many other technologies. Furthermore, a high-numerical aperture (high-NA) extreme ultraviolet center in the complex, which includes federal funding, will allow exploration smaller than 1 nanometer (sub-1 nm).

Turning to a discussion on manufacturing, Johnson explained that the peak year for manufacturing employment in the United States was 1978, with a systematic decline through 2010. However, he emphasized that manufacturing “value-add” is succeeding; the United States has been improving continuously, with $3 trillion of manufacturing output per year over the past 25 years. During the same period, Japan’s value-add decreased, Germany’s flattened, and China’s steadily increased, with China having the largest manufacturing value-add in the world. He pointed out that the United States is also excelling in terms of manufacturing per person in the economy; however, U.S. manufacturing is not uniformly distributed across sectors or geographically, and manufacturing comprises less than one-third of the U.S. economy. South Carolina, in particular, is a manufacturing-based economy; 70 percent of cars made there are sold overseas, which demonstrates the United States’ competitive potential. In fact, U.S. states compete vigorously to locate factories in their jurisdictions. To move forward, he suggested studying examples of public–private partnerships and considering how best to adapt them. For example, current U.S. partnerships are based either on technology assets (e.g., national laboratories, federally funded research and development centers [FFRDCs], and the Manufacturing USA institutes) or human talent (e.g., land-grant universities and 2-year colleges). Regarding the former, he advocated for creating better open-access environments and thinking beyond national boundaries, especially in terms of strategies to share data in supply chains. For the latter, he stressed that people are the key to keeping semiconductor equipment running and the mid-scale manufacturing industry moving forward, and increased investments in students as well as in existing public–private partnerships are critical.

Sharing her insights on data-informed manufacturing, Lambert remarked that supply chain data can be misleading, and the simplest data often are the most useful. Reflecting on pharmaceutical supply chain data in particular, she said that records of all legal imports to and exports from the United States are maintained by the U.S. Census Bureau in USA Trade Online.⁹ These data are free, publicly available, anonymized, and organized by a commodity code based on the harmonized tariff schedule. For example, to better understand a recent concern about a potential penicillin shortage, using USA Trade Online, she discovered that penicillin was previously imported from Spain. When that factory closed, the United States never made new import plans and other suppliers could not meet the demand. With this example in mind, she

highlighted opportunities to request more data from the federal government at no cost, which can be helpful for those determining whether to make initial investments in manufacturing. These data can be requested by asking, in writing, for a new commodity code. For instance, after making a request for new data on amoxicillin for oral suspension, she learned that most amoxicillin for oral suspension is imported to the United States from Jordan. She asserted that such data can be used to “foresee the foreseeable and avoid the avoidable,” in terms of pricing trends, tariff reactions, sanctions lists, and natural disaster areas. In addition to USA Trade Online for U.S. trade data, the United Nations’ database Comtrade¹⁰ is useful for international trade data, and the U.S. Bureau of Labor Statistics has data on education and training. For critical materials in particular, she pointed out that data can help illustrate from where the United States imports and the cost per unit of these imports from each country.

Discussion

Wojtecki posed a question about managing IP in an open environment. Kelly explained that establishing principles is key to avoiding any IP leakage or contention. In the case of NY CREATES, he said that a “hub and spoke” model is used to manage IP—the IP remains with the industry partners, as the state is only interested in economic development. Johnson noted that at universities, IP activity primarily occurs when students graduate and transfer knowledge to the private sector. He agreed that developing fair and transparent rules, as well as having standard term contracts in place, is necessary in partnerships. Wojtecki asked whether IP presents restrictions related to commodity codes. Lambert replied that according to the U.S. Census Bureau, in order to maintain anonymity and avoid revealing trade secrets, codes cannot be made for a commodity unless it has at least three importers/exporters and at least $1 million in import/export.

Hattrick-Simpers wondered how to take better advantage of data to develop and evolve training programs. Johnson emphasized the value of training people with skills that are transferable instead of training people to work only in a specific factory. He championed BMW’s and the U.S. Air Force’s approaches to training and he suggested a strategy that both considers existing manufacturing assets and a core set of transferrable skills and adopts augmented reality and advanced robotics to train people faster than competitors. Lambert highlighted statistics on the labor force that categorize the skills required for different jobs, which help businesses make decisions about training programs. Johnson pointed to the Federal Reserve Economic Data as another valuable data resource on manufacturing wages. He stressed that one of the best ways to recruit more workers is to pay them more.

Morgana Trexler, JHU APL, posed a question about what could be learned from other countries that excel in manufacturing. Reflecting on Ireland’s progress in particular, Johnson noted that modernizing capital infrastructure is key. Furthermore, he said that Switzerland’s exceptional education system leads to the manufacturing of high-value-add products. He stressed that the United States can learn from others, although it has different challenges related to scale and legacy infrastructure in a diverse economy.

Kingon inquired about public–private partnerships and productivity. Kelly said that NY CREATES is driving down the cost per transistor at an incredible rate. However, companies that did not leverage the Complex were not able to advance at that rate and fell off the productivity curve. Johnson added that productivity connects to the issue of human capital, with consideration back to the value of K–12 science and mathematics courses as well as training without creating debt for students. He urged continued investments in both people and technologies.

Vaia asked what motivated the level of collaboration and innovation of NY CREATES. Kelly explained that IBM used to build tools, but that approach was too expensive and unsustainable, especially without the best talent in the world. IBM then began developing partnerships with companies to create successful tools and the industry was carved into platform companies at different layers—the Albany NanoTech Complex is an extension of that. Kelly also said that he championed opportunities for DoD to build a similar structure as the semiconductor industry did to develop the Manufacturing USA institutes, when

she championed agile, reconfigurable manufacturing to enable rapid adaptation to new product designs; modular, distributed manufacturing sites that work in concert to ensure rapid scaling and high yield; and enhanced quality and performance.

Discussion

In response to a question from Jian Cao, Northwestern University, about APL’s funding, Hamilton explained that almost all funds are from DoD, but collaboration with industry is possible in some cases. She added that APL develops and protects IP but transitions licenses to the government for use. Basceri inquired as to whether any surprises occur when transitioning an innovation to production. Hamilton stressed just how much time it takes to translate a laboratory development to industry but said that APL’s technology transfer office helps with the transfer of IP.

Kurfess posed a question about filling gaps and scaling to higher (MRLs). Hamilton commented that APL has the ability to scale up to a certain degree, approaching what mid-scale manufacturing might look like, to better understand relevant problems. She asserted that APL makes the most progress when it understands the risks in the laboratory and then collaborates with an agile industry partner with resources for experimentation—a type of collaboration that helps bridge the valley of death.

Johnson wondered whether recent AI funding launches new opportunities in materials science with which computer scientists could engage. Hamilton indicated that APL has teams with data scientists, materials scientists, and computational experts. She noted that it takes time for all team members to “speak the same language,” but APL seeks new collaborations to advance both materials science and algorithm development.

Cao asked Hamilton about high-throughput testing for materials discovery. Hamilton replied that directed energy deposition is only appropriate in certain situations, such as the lunar materials project. She advocated for mimicking the processing conditions that will be most realistic for a desired material and incorporating variation in processing conditions when optimizing for synthesis. She added that the United States has developed successful demonstrations of flow chemistry and additive manufacturing for metals but is moving toward more versatile high-throughput, self-driving laboratories and fabrication facilities. Trexler inquired about the most disruptive emerging technology for mid-scale manufacturing, and Hamilton expressed her excitement about high-flow, high-throughput chemistries and novel biomanufacturing techniques.

Another participant asked whether APL is concerned about byproducts of or gases in the chemical processes in its recycling projects. Hamilton acknowledged that although this is a valid concern, APL has learned important lessons from previous process developments and no concerning byproducts have emerged from the gallium alloy.

INNOVATION AND INTELLECTUAL PROPERTY

Conley emphasized that the scaling challenges of prior generations remain today. Using the F-35 weapons platform as an example, he noted that IP helps drive innovations to support DoD, but IP is also a “heartburn issue.” Because the IP for this platform was never transferred from Lockheed Martin, many issues arose when the associated software did not work well. This problem is not unique to DoD, however. Farmers who own John Deere tractors have to use expensive authorized dealers for repairs because they do not own the tractors’ software. He suggested that novel contracting techniques could address these types of issues.

Quoting Abraham Lincoln, Conley elaborated that “the patent system add[s] the fuel of interest to the fire of genius.” As a way to better showcase the “fuel” (e.g., strategic partners or investors) and the “fire” (e.g., the IP), he introduced a case study of a disruptive energetics technology for rockets brought forward by a PhD student at Purdue University. The patent for this solid rocket propellant was owned by the university, but a few years later, the student refined the proprietary nature of the fuel and the patent for that technology is now owned by Adranos. A little more than 1 year later, a strategic acquisition occurred, and Anduril acquired Adranos, which led to increased investments from venture capitalists and more sources of human capital.

Conley suggested using Crunchbase to better understand this type of investing activity and to consider potential

opportunities for mid-scale manufacturing. He also championed the use of the Patent Asset Index to compare firms using the sum of technology relevance (based on citation count adjusted for patent office practices, age, and field) and market size. Conley also suggested that this methodology can be used to learn more about a company’s value by understanding its IP, and related technology clusters can display the specific contents of a patent portfolio to better analyze investors, inventors, and patent value.

Conley explained that supply chain analytics are also evolving to capture production- and demand-related data efficiently, which might help with demand forecasting as well as illuminate partnerships and IP. As an example, he posited that aggregate demand across munitions and other DoD programs will cause long-range anti-ship missile (LRASM) production capacities to reach a limit. For example, conflicts with near-peer adversaries would drive large increases in demand for LRASM as well as adjacent mission-critical munitions. In addition, increases in production beyond current budget forecasts are necessary to provide adequate stockpiles of munitions and backfill depleted supplies. Limited analysis has been done in the past to map the supply base and determine the feasibility of accommodating large demand surges across multiple programs that share a common industrial base and the impact of foreign dependencies. To address these risks, BlueVoyant Government Solutions¹¹ developed a network assessment of the LRASM industrial base by analyzing the supply case, developing a taxonomy of key systems and mapping suppliers to the systems they support, measuring centrality to identify critical “chokepoint” suppliers, and establishing linkages to other programs and sources of supply to identify bottlenecks. To summarize, he suggested that IP and supply chain analytics could merge to support dual-use defense contracting and demand signal management.

Discussion

Vaia acknowledged that a high number of patents does not necessarily correlate with an organization’s level of innovation, and he inquired about specific uses of the Patent Asset Index. As an example, Conley said that this methodology has been used to determine which firms in which industries were moving toward patent portfolios that supported the United Nations’ Sustainable Development Goals. In another instance, Finland used the methodology to learn which firms’ portfolios were moving away from carbon-based fuels and rewarded those entities with government funding.

Basceri asked whether the Patent Asset Index can measure the amount of “know-how” that is behind (but not part of) a patent. Conley explained that for large patent portfolios, abundant trade secrets and know-how support production, which creates a problem for an analysis of the supply chain. Because people do not want to share information about the quality control in their supply chain, he said that this can only be intuited based on the number of suppliers. Gamota described an approach of publishing in obscure journals instead of patenting certain information. Manufacturers also might embed know-how—nuances that make a difference in terms of quality and yield. He said that if one could identify where these “secret sauces” reside, knowledge would expand. On a related note, Kingon advocated for helping PhD researchers to develop better research skills, learn to use databases and other analytic capabilities (beyond Google Scholar and ChatGPT), and learn how to link analyses.

Cao commented on the limited knowledge that small businesses in a supply chain might have. Conley said that this is not a problem for DoD (unless the material is secret) and DoD could use these transparent analytic tools to encourage dual-use opportunities among suppliers. Hattrick-Simpers cautioned that DoD has to avoid any appearances of endorsement and conflict of interest.

SMART MANUFACTURING

Kurfess explained that smart manufacturing could transform the U.S. manufacturing sector. The future of U.S. manufacturing is envisioned to move from isolated, optimized cells to fully integrated data and product flows across borders, with enhanced communication, greater automation, and machine–machine and machine–human interaction. He asserted that next-generation technologies such as AI and ML, high-speed connectivity, advanced data

analytics, and automation could improve productivity, efficiency, and sustainability for the manufacturing workforce, factories, and supply chains; expand the smart manufacturing workforce; and increase U.S. economic competitiveness and resilience.

Kurfess presented five key recommendations from the National Academies’ 2024 report Options for a National Plan for Smart Manufacturing¹²:

1. “A national plan for smart manufacturing should urgently support the establishment of national transformative data infrastructure, tools, and mechanisms to assist with (1) cultivating, selectively sharing, and securing the use of data in real time and at scale; and (2) sharing best practices to promote industry-wide technical data standards. Such infrastructure could take the form of a secure digital network that facilitates the flow of data with controlled and credentialed access, such as a Cyber Interstate. It should be planned and coordinated with companies, government agencies, associations and consortia, and academic stakeholders.”
2. “The Department of Energy in partnership with the National Institute of Standards and Technology, the Department of Defense, and manufacturing institutes should establish manufacturing CASE (Calibration, Autonomy, Security, Evaluation) Data Banks with the next generation of secure manufacturing architectures.”
3. “Smart manufacturing is multifaceted, and technologies developed in one specialty area most likely will not be suited for other applications. The Department of Energy and other federal agencies should fund programs and consortia that develop technologies at the intersections of critical technologies (e.g., human–AI copiloting, sensing, AI/machine learning, platform technologies, digital twins, uncertainty quantification), unit manufacturing processes (e.g., casting, forming, molding, subtractive, additive, and joining), and industry sectors (e.g., semiconductor, aerospace, automotive, biomedical, and agriculture).”
4. “Funded by the Department of Energy, in consultation with other relevant federal departments and agencies, a framework should be developed to quantify the broader sustainability benefits of implementing secure smart manufacturing (considering three pillars: environment, economy, and society) as well as industry-wide sustainability metrics.”

In addition to these five key recommendations, Kurfess noted that the report made several other recommendations, highlighting the need to assess and invest in education for the smart manufacturing workforce; support workforce development with financial incentives and learning opportunities; provide career support via fellowship programs, continuing education programs, career exploration opportunities, tax incentives, wraparound services, and more; develop smart manufacturing programs such as pilot programs for data sharing, resources for small-to-medium manufacturers, and access to expertise and facilities; and increase strategic coordination among federal agencies.

Kurfess also shared some of the report’s conclusions, indicating that because smart manufacturing is “horizontal and crosscutting,” industry and government strategies are essential. Public–private partnerships are vital for the future of U.S. smart manufacturing, he continued, and the Manufacturing USA institutes in particular could play an important role in advancing the industry. He emphasized that this national plan for smart manufacturing could “disrupt current industry practice, stimulate new market drivers, motivate new ways of thinking about data, and drive more rapid approaches to addressing environment stability.” By elevating the manufacturing workforce and ecosystem, as well as increasing investments, cutting-edge capabilities could expand in all manufacturing sectors and help to safeguard the U.S. economy and national security.

Discussion

Gamota asked whether the National Academies’ report considered specifically how data could be integrated, anonymized, and shared. Kurfess explained that access could be controlled at different levels to enable sharing, and Cao added that the phrase “data bank” was carefully chosen (versus “database”) because it means that data can be viewed as a “currency” that can be “traded.”

Conley indicated that if a national smart manufacturing plan supports proximity (i.e., the source of the supply moves closer to the moment of demand), the private sector might be more willing to invest. Johnson added that strategies are needed to ensure that the United States is an active player among other countries, many of which already have national smart manufacturing policies.

Vaia pointed out that although manufacturing initiatives represent added value, new entrants to this market will be essentially “taxed.” Costs will surface to maintain a data bank, he continued, and a sustainment model for implementation is needed at the federal level. Kurfess noted that the groundwork for implementation has been laid. Jennifer Wolk, Office of Naval Research, posed a follow-up question about embracing smart manufacturing at the small- to mid-scale levels without “crippling costs.” Because the defense sector prioritizes protecting both data and technology, she also wondered about the balance needed to address mid-scale gaps. Lambert commented that because agencies under the same secretary were not even willing to share data during the COVID-19 pandemic, people wasted valuable time and effort requesting the same information from the same companies, thus crippling them. Therefore, she encouraged DoD to determine how to take advantage of data already being gathered, avoid immobilizing businesses in the market, and enhance business propositions.

MANUFACTURING EFFORTS IN ENERGETIC MATERIALS RESEARCH

David Bahr, Purdue University, described the Purdue School of Materials Engineering, which has 36 full-time faculty and laboratories in 10 buildings across four zip codes and two area codes, including a heat-treating consortium, a surface engineering and enhancement consortium, a laboratory for advanced materials and processing, and the Purdue Energetics Research Center (PERC).¹³ PERC is a cross-college effort in energetics research that requires substantial investment from the university and external partners to “make more than a gram” of material to better understand real-world phenomena. One hundred forty students, faculty, and staff participate, with $90 million in open energetic materials contracts at present.

Bahr defined energetic materials as those that do not require an oxidizer, that have temperature-sensitive reactions, and that are highly exothermic. They can be used to make explosives, propellants, and pyrotechnics. He explained that PERC’s foci include synthesizing new energetic molecules, designing and fabricating new and existing formulations using new materials and methods, modeling energetic materials and formulation behavior at the mesoscale and macroscale, performing characterization and testing of energetic materials and formulations, performing optical diagnostics of post-detonation fireballs, evaluating high-rate mechanics, designing and performing bulk and additive manufacturing of energetic materials and formulations, and modeling manufacturing risk and sustainability.

Bahr described a sample project from PERC in energetic materials synthesis, which involved making new molecules from nitrogen and carbon strains; when those bonds break, energy that translates to heat is released, which could be used for detonation or as a propellant. PERC also makes common explosives such as Royal Demolition eXplosive (RDX), CL-20, and TKX-50, for example, in addition to conductive binders and metallized propellants—with standard operating procedures and safety checks a key part of any process. Students participate in extensive training to develop these materials, making the program a critical part of workforce development. He stressed that Purdue’s product is not data but rather people. Returning to the example of the student-developed solid rocket propellant discussed by Conley—Bahr emphasized that Purdue does not just want students to publish papers—that student explored a specific interest in commercialization, learning about IP, taking extra courses, and ultimately successfully transitioning the technology from Purdue to Adranos.

Bahr also discussed PERC’s role in additive manufacturing of energetics, with a particular interest in vibration-assisted printing of highly viscous propellants. The goal of this project was to put as much energetic material into a binder in three-dimensional printing as possible. He added that students first worked with mock materials (e.g., food products, pharmaceuticals) before

loading with energetics to reduce the risk and cost of experimenting with dangerous explosives.

Bahr underscored that working only on the small scale is insufficient; having some manufacturing capacity in-house to scale is critical. He indicated that, previously, PERC worked with pharmaceutical companies to build a pilot plant for tableting to produce a real volume of pharmaceuticals. In a similar vein, PERC is developing a small-scale fabrication facility for energetics to make higher volumes of materials for use by other customers.

Discussion

Gamota posed a question about PERC’s decision-making processes. Bahr replied that no projects are initiated without faculty interest and all projects are run by faculty. Each group makes decisions about which parts of the research to pursue and which tools to buy and use; because this process is not streamlined, he said that additional scalability challenges can arise.

Vaia observed the value that PERC’s small-scale manufacturing plant offers in training students but wondered how the university manages that entity to help small companies scale. Bahr noted that Purdue has an applied research institute that is a secure facility with primarily staff workers contracted to deliver products on a small scale. Guardrails are in place to avoid direct competition with industry and Purdue charges competitive rates so that mid-sized businesses are not disadvantaged. Another participant asked about the value of Small Business Innovation Research and Small Business Technology Transfer (STTR) programs. Bahr responded that although they are good tools, other tools might be more appropriate in certain cases.

Joe Poshusta, Synthio Chemicals, inquired about successful collaborations between industry and universities more broadly. Bahr noted that initial relationship-building is challenging, but once a partnership is established, it is sustainable. Universities are usually decentralized, he continued, with various groups working on energetics with different perspectives, goals, and timescales; problems can arise when industry has to communicate with too many layers of people at universities. In response to a follow-up question about how government could better initiate partnerships, Bahr suggested ensuring that the initial contractor understands what guardrails are already in place at a university.

Johnson remarked that, in general, the manufacturing sector needs more people. He encouraged participants to review the Cleveland Federal Reserve’s work on the wage premium in the manufacturing sector, which is heavily weighted toward leadership jobs, not critical production jobs.

ENERGETICS: UNIQUE FEATURES AND AN ECOSYSTEM FOR SCALE-UP

Keith Whitener, DARPA, discussed five challenges related to energetics and potential technological strategies to address them. First, safety in manufacture, transport, and deployment is a key concern for those entering the energetics market. Many regulators are (necessarily) involved, which might discourage smaller firms from engaging. He explained that safer energetics could be made—for example, DARPA’s SeREne program¹⁴ is exploring engineering energetics at the materials scale so that they are safe until they need to perform. However, to convince regulators that energetics are safer, better metrology would be needed, as investigated by the DARPA RIDE program.¹⁵ He noted that remote or autonomous systems could also improve the safety of energetics manufacturing.

Second, Whitener indicated that the potential toxicity of compounds and concerns about releases to the environment should be considered, and Environmental Protection Agency involvement, although also necessary, could discourage more firms from entering the market. He emphasized that trade-offs are essential in this case, as introducing new properties can degrade energetics performance.

Third, Whitener noted that security is paramount in energetics manufacturing, and data sharing is limited. Accordingly, establishing security clearance for all employees is a time-consuming and expensive process. This, too, might discourage individuals from working in the field. Furthermore, he said that manufacturing of energetics and munitions is highly centralized; when only one plant in the United States makes the explosive 1,3,5-trinitro-1,3,5-triazine or RDX, single points of failure can emerge. He remarked that compartmentalization works well to address security concerns with remote or autonomous systems while also reducing the number of people that need security clearance. He added that regulatory systems are also needed to engage small businesses in controlled or classified projects.

Fourth, Whitener said that the range of energetics products is vast, which creates issues in workforce development and the transferability of specialized knowledge. Although the value of addressing this issue with an autonomous distributed manufacturing system remains to be seen, he noted that standardization and modularity to stockpile better-aging components is a potential solution.

Fifth, Whitener described the fragility of the energetics supply chain: if one component of the supply chain fails, or if an adversary contributes counterfeit parts, national security could be at stake. Furthermore, he noted that these “wildly complex technological marbles” are single use, which affects supply and demand, and an aging stockpile might not be functional when needed. If millions of rounds of munitions are needed, an issue could arise with manufacturing and quality assurance. He stressed that prioritizing production of new munitions and platforms is key, as is reshoring feedstocks and explosives such as trinitrotoluene (TNT). Digital twin quality assurance approaches could also be helpful, and distributed manufacturing could address issues with stockpiling and quality control.

However, Whitener pointed out that while these are not insurmountable obstacles from a technological perspective, economics plays a key role. Although he said that the government likely is the only buyer of munitions, this means that the demand is uneven, and planning for the future at the individual company level is difficult. Public–private partnerships could be beneficial, he continued, or the United States could decide that resilience in energetics is valuable and pay a premium for it. New entrants could then launch a distributed manufacturing point of production and receive incentives. He suggested a quasi-dividend system where money put into the energetics community is redistributed.

Discussion

Johnson emphasized that return on investment has to be considered for all potential solutions. He also pointed out that companies have a choice whether to develop products for the government or another customer, which could be problematic as products become more expensive and companies have to make economic decisions about where they invest. He added that U.S. resilience is a challenge; because maintaining a surplus is not feasible, the United States has to determine how it will flex quickly when needed. Johnson and Whitener agreed that case studies of potential business models would be valuable. Conley introduced a model in which large funders who build a plant in the United States would not have to pay tax on their return on investment for a certain period of time. Kingon remarked that unique economic incentives are needed to entice new players into the market while minimizing additional cost to the government. Kurfess said that the National Academies’ report on smart manufacturing recommended tax breaks for companies upgrading equipment and investing in technology, which could support fast-paced innovation. Reflecting on other possible financial incentives, Johnson observed that R1 institutions cannot survive without international students. He wondered about strategies to attract more U.S. students (e.g., GI bill students) by providing financial supplements for tuition and to maintain credentials.

PANEL ON ELECTRONICS AND CHEMISTRY

Poshusta described Synthio as “a case study in mid-scale manufacturing.” Founded in 2017, the small company focuses on integrated processes, automation, hazardous reactions, and high mass transfer rates. He explained that Synthio works with exothermic reactions and hazardous feeds or products at a scale of tens of tons per year of production, enabling a safe work environment with a competitive advantage. Reflecting on the dual development cycle for energetic and critical materials for defense,

integrated. The resulting solution, known as S90LN (Low Noise), delivers enhanced readout performance for thermal imaging applications and is now available as part of the recently launched ThermaView™ Solutions, further demonstrating SkyWater’s ability to adapt and scale differentiated technologies rapidly. In closing, he encouraged programs to dedicate more attention to metrology and test capabilities, which are critical to generate new knowledge during product development and to increase manufacturing efficiency. He mentioned that although the CHIPS and Science Act of 2022 has benefited the semiconductor industry overall, only $2 billion has gone to mature-node facilities like SkyWater and opportunities are being missed.

Discussion

Hattrick-Simpers posed a question about AI readiness. Poshusta said that advanced online sensors could be used to run steady-state processes, but integration is challenging, and feedback is needed before AI can be used to help control processes. As part of its partnership with Purdue, Synthio is exploring how to maintain and reuse software in new processes or to enable a partner’s compatibility with Synthio’s system. Woodruff noted that systems that gather data and workflows that present information are needed as well as systems that can react more autonomously. EMD Electronics’s automation activities focus on modularity and orchestration (e.g., whether AI can provide step-by-step instruction to a piece of hardware clearly and safely). Korolchuk added that SkyWater would benefit greatly from the application of AI and the centralization of data. He pointed out that advanced software systems exist in certain verticals, but these systems are from different software vendors, which leads to fragmented solutions.

Cao inquired about lessons learned on data sharing as well as the value of a centralized data entity. Woodruff explained that the proof of success is an improvement in a customer’s yield or performance. Because of AI’s popularity, he said that companies realize the value of revolutionizing how they handle data, although additional leadership is key to success. Explaining the benefits of data sharing might motivate people to share more data, he continued, but a trusted interface is essential.

Hattrick-Simpers expressed concern that new methods of optimization that are found at the fab level are not available at the fabless level. Korolchuk validated this concern, describing a recent effort to improve performance via non-Manhattan geometry for lithography, which requires collaboration between fab and design. Woodruff described the trend in chip design improvements to accelerate AI compute performance that augments classic Moore’s Law, and as such CMOS scaling performance improvements, but he noted that different materials and device architectures are needed to enable continued progress in AI compute performance and efficiency. He stated that achieving repeatable, reliable processes takes much time and effort.

Wojtecki posed a question about modernizing to remain at the forefront of capabilities and to attract more customers. Korolchuk said that SkyWater analyzed a matrix of needs across a wide range of customers and technology trends and intends to work with vendors to develop wafer-level automated radiation testing and cryogenic automated testing as well as to make those capabilities accessible at a lower cost. Basceri asked how SkyWater will differentiate its business in such a competitive foundry space, and Korolchuk replied that the company is not interested in competing with global foundries that better serve very high-volume customers. It differentiates its technology capabilities for its partners and is always seeking new technologies that are compatible with its manufacturing capabilities. Bahr inquired about plans for maintaining the capacity of older-generation equipment. Korolchuk explained that SkyWater communicates actively with its equipment vendors that publish timelines on the support of legacy equipment. SkyWater then modulates its capital spending based on those suppliers’ plans.

Basceri asked about Synthio’s investors and Poshusta replied that the company is mostly privately held with a few external investors; DoD has also invested a substantial amount. Given the risk about whether orders for certain materials will emerge, Synthio hopes to make its technology reusable in a variety of applications and to address other markets. Vaia inquired about the advantages and challenges of federal versus private investments. Poshusta said that Synthio’s government funding is largely nondilutive and Synthio owns much of the IP. Although those are valuable investments, he continued, they might

not be as aligned with Synthio’s business goals as those from private strategic investors. He also expressed the need for greater clarity on demand-scale when selling products to government contractors. Woodruff added that ensuring investments will pay off is key and regular engagement with partners helps align prioritization of the work. Korolchuk explained that SkyWater values a balanced approach. He said that DoD relationships generally have been positive experiences, but some DoD programs have criteria that make it difficult to serve other DoD programs. Dan Cole, Army Research Office, posed a question about how companies assess whether to invest their own funds in new materials. Woodruff remarked that this decision is influenced by a technology’s potential to enable superior performance. EMD Electronics considers work in academia and trends in compute; if government agencies can involve relevant customers, validity and motivation increase.

Wojtecki asked about the regulatory environment for energetic materials. Poshusta described it as “a maze and a complicated environment.” He explained that different agencies regulate from safety, transportation, or environmental perspectives; sometimes the framework is clear, but other times it is unclear or overlap exists. More flexibility among the regulatory agencies in terms of updates would be beneficial. Scale is another challenge, he continued. Synthio tends to work with materials as dilute solutions (i.e., not explosive), which helps gain some quantity, but to explore larger-scale production, concerns about atom efficiency and environmental controls arise.

WORKSHOP TAKEAWAYS

Workshop planning committee members highlighted important workshop takeaways, such as for leadership from industry to coordinate partnerships more effectively, both technologically and economically. Some stressed that mid-scale manufacturing needs to make economic sense to advance. Although AI could be a “game changer,” fundamentals and know-how remain critical to advance technologies; facilities cannot be operated without people with the right skill sets.

Another member next shared five key themes observed during the workshop. First, the government can take different roles and can shape a field. Government funding, including public–private partnerships, influences the manufacturing sector as well as how technical potential is identified and investments are deemed worthy of risk through either public good or for-profit dual use. Furthermore, single buyers and single suppliers both create challenges and more buyer interest could be stimulated with distributed manufacturing and smart manufacturing. Additionally, increased efforts to derisk investments could entice more suppliers into the supply chain. Workforce issues also arose during the discussion of this theme, highlighting human capital as the value of university research. Second, access to capital, equipment, and infrastructure is critical for scaling, but questions remain about the best models and the best strategies to secure funding. Third, implementing autonomous materials discovery could accelerate innovation and could be linked to development needs and available supply chains. Fourth, different perspectives exist on how to use IP, how IP can be mined to understand companies’ progress, and how one can track government investments through IP. Fifth, incentives are valuable and opportunities exist for the government to create them.

Disclaimer: This Proceedings of a Workshop—in Brief was prepared by Linda Casola as a factual summary of what occurred at the workshop. The statements made are those of the rapporteur or individual workshop participants and do not necessarily represent the views of all workshop participants; the planning committee; or the National Academies of Sciences, Engineering, and Medicine.

Workshop Planning Committee:Julie A. Christodoulou (Chair), Office of Naval Research (retired); David E. Aspnes (NAS), North Carolina State University; Jian Cao (NAE), Northwestern University; Jason R. Hattrick-Simpers, University of Toronto; Angus Kingon, Brown University; Thomas R. Kurfess (NAE), Georgia Institute of Technology; Morgana Trexler, Johns Hopkins University Applied Physics Laboratory; Rudy Wojtecki, Applied Materials; and Pablo D. Zavattieri, Purdue University. The National Academies’ planning committees are solely responsible for organizing the workshop, identifying topics, and choosing speakers. Responsibility for the final content rests entirely with the rapporteur and the National Academies.

Reviewers: To ensure that it meets institutional standards for quality and objectivity, this Proceedings of a Workshop—in Brief was reviewed by Krystel Castillo-Villar, Department of Energy Cyber Manufacturing Innovation Institute; Leila Ladani, Arizona State University; Jen Dailey Lambert, Johns Hopkins University Applied Physics Laboratory; Rudy Wojtecki, Applied Materials; and Katiria Ortiz, National Academies of Sciences, Engineering, and Medicine, served as the review coordinator.

Sponsor: This Proceedings of a Workshop—in Brief is based on work that was sponsored by the Army Research Office and was accomplished under Grant Number W911NF-23-1-0409. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the Army Research Office or the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation herein. Any opinions, findings, conclusions, or recommendations expressed in this publication do not necessarily reflect the views of any organization or agency that provided support for the project.

Staff:Erik Svedberg, Samantha Koretsky, Maddi Nicol, Amisha Jinandra, Joseph Palmer, and Michelle Schwalbe

Suggested citation: National Academies of Sciences, Engineering, and Medicine. 2025. Mid-Scale Manufacturing and Characterization Capacity for Department of Defense Critical Materials Supply Challenges, Part 1: Proceedings of a Workshop—in Brief. Washington, DC: National Academies Press. https://doi.org/10.17226/29117.

For additional information regarding the workshop, visit https://www.nationalacademies.org/our-work/mid-scale-manufacturing-characterization-capacity-accelerating-scale-up-and-transition-of-technologies-to-mitigate-dod-critical-materials-supply-challenges-a-workshop.

Division on Engineering and Physical Sciences

Copyright 2025 by the National Academy of Sciences. All rights reserved.