Cold Fusion & LENR: Navy & NASA Investigations

October 30, 2026
|In Energy
|By Tim Ventura

Three Navy laboratories, NASA experiments with hydrogen-loaded metals, a $10 million Department of Energy initiative, and a DARPA program targeting fusion inside solids: cold fusion’s institutional history is more substantial—and more complicated—than its reputation suggests. In a conversation with Dr. David Nagel, a former Naval Research Laboratory division superintendent, that history becomes an engineering story about materials, measurements, and reproducibility.

A Navy Scientist’s Unexpected LENR Assignment

David Nagel’s involvement began with a tap on the head. On March 24, 1989, he was swimming laps at the Naval Research Laboratory when a secretary interrupted him: an associate director wanted to see him immediately. Nagel hadn’t heard about the Fleischmann–Pons announcement the previous day, but the questions it raised fell directly within his responsibilities in solid-state and nuclear physics. He quickly organized a meeting. Roughly three dozen people arrived from physics, chemistry, and materials research, and the laboratory began examining a claim that was already challenging the scientific establishment.

Nagel wasn’t approaching the problem from the margins of institutional science. The Navy had assigned him to NRL in 1962 while he was on active duty. He subsequently stayed as a civilian physicist and eventually managed an organization that, by his account, included approximately 150 people, 80 PhDs, and an annual budget equivalent to about $60 million in today’s dollars. Those figures describe his broader research division, not a dedicated cold-fusion program. In 1998, he retired, took a weekend off, and started at George Washington University. At 88, he’s still conducting experiments and thinking about how materials could enable new energy technologies.

His confidence in LENR—low-energy nuclear reactions—didn’t arrive with the first announcement. Nagel says it developed during the mid-1990s, after he’d examined experiments, learned the details of calorimetry, and worked through questions about calibration and interpretation. He readily acknowledges that researchers contributed to the early controversy by announcing effects that sometimes turned out to be detector noise. Scientific objections were real, and some experiments failed. His position is that those difficulties justified better investigation, not the permanent abandonment of the subject.

That distinction shaped our conversation. Nagel is convinced that the accumulated evidence warrants an engineering effort, but he doesn’t confuse that conviction with possession of a finished generator. When I asked whether cold fusion could be engineered, he answered that, from his perspective, it both could and must be. The work still required is less glamorous than the original promise of limitless energy: identify what makes a material active, measure what it produces, understand why it changes, and establish a process that another laboratory can reproduce.

Three Navy Laboratories, One Reproducibility Problem

The Office of Naval Research brought three laboratories into a coordinated program beginning in January 1992. At the Naval Research Laboratory in Washington, D.C., investigators worked on palladium preparation, material characterization, hydrogen-isotope loading, calorimetry, and radiation measurements. One diagnostic used changes in a cathode’s electrical resistance to estimate how much hydrogen or deuterium it contained. That mattered because buying palladium and immersing it in an electrolyte didn’t guarantee that two laboratories had created the same internal conditions.

At the Naval Air Warfare Center Weapons Division in China Lake, California, electrochemist Melvin Miles and his collaborators investigated palladium electrodes in heavy-water systems. Their distinctive contribution was studying reported excess heat alongside helium production. A calorimeter measures energy flow; gas analysis looks for a possible nuclear product. Establishing a relationship between those measurements would be more informative than observing a temperature rise alone. Nagel remembers Miles as a particularly capable experimentalist whose work helped sustain the Navy’s interest.

The San Diego group, associated with the Naval Command, Control and Ocean Surveillance Center’s research division, NCCOSC–NRaD, and later SPAWAR, pursued another approach. Researchers including Stanislaw Szpak and Pamela Mosier-Boss investigated palladium–deuterium co-deposition: forming the palladium deposit while introducing deuterium. Rather than treating the electrode as a finished component whose structure was fixed before the experiment, they made material formation part of the experimental process. Their investigations encompassed thermal behavior and reported nuclear emissions, with the evolving deposit and its environment central to the interpretation.

Nagel describes the program as complementary rather than uniform. Most experiments involved electrochemistry, but NRL also investigated gas-phase approaches. Researchers visited one another’s laboratories, compared methods, and argued about results. “Normal science,” he called it. Yet he also made an important concession: the program didn’t achieve the interlaboratory reproducibility its managers wanted. The three-laboratory history demonstrates sustained institutional investigation, not three identical confirmations of a settled phenomenon. Understanding the differences between their results remains part of the scientific problem.

The Numbers Behind the Cold Fusion Dispute

China Lake’s 1996 report, Anomalous Effects in Deuterated Systems, provides concrete measurements behind the controversy. Its authors described typical excess-power fractions of 5–10%, a maximum of 30%, and reported values reaching 0.52 watts of excess power and 1,400 kilojoules of excess enthalpy. Power and accumulated energy aren’t interchangeable: a small persistent imbalance can become significant over a long experiment. These were the researchers’ reported findings, however, not specifications for a validated energy product.

The same report makes the reproducibility problem numerical. Only about 30% of the experiments overall reportedly showed excess power. Results varied with the palladium: seven of eight experiments using palladium–boron cathodes were positive, compared with 17 of 28 using one commercial source. The authors also described initiation periods of approximately 6–14 days and a current-density threshold around 100 milliamperes per square centimeter. Their argument was that the material’s source and processing history weren’t incidental details; they could determine whether anything happened.

NRL’s January 1996 report reached a different experimental outcome. It reported no excess power above 200 milliwatts in the specified electrolytic tests and no anomalous radiation in the measurements described. In a later introduction, Miles argued that changes to the calorimetry had increased uncertainty from approximately 20 milliwatts in his apparatus to approximately 200 milliwatts at NRL, potentially obscuring effects he usually measured. That criticism doesn’t erase the null results. It shows why replication disputes can turn on heat losses, instrument sensitivity, and modifications that initially appear to be improvements.

For Nagel, the strongest evidence comes from combining measurements that have different failure modes. He emphasizes heat–helium correlations because calorimetry and mass spectrometry involve different instruments, calibrations, and analyses. Their agreement can strengthen a case, although it doesn’t remove the need to exclude contamination, backgrounds, and systematic errors. The broader lesson is methodological: an experiment needs to account for its energy inputs and outputs while independently examining possible reaction products. That’s also why newer federal programs place such emphasis on measurement capabilities.

NASA Pursues Lattice Confinement Fusion

NASA’s involvement also extends well beyond a single experiment. A 2012 NASA Glenn retrospective described limited investigations dating to 1989, including loading palladium with hydrogen or deuterium gas and exposing palladium films to multibubble sonoluminescence. Nagel noted that both NASA Glenn and NASA Langley had worked in the broader area. These weren’t all versions of the same apparatus, and their objectives ranged from investigating temperature effects to examining nuclear processes. The agency’s later lattice-confinement work developed one particular route through that larger experimental landscape.

The 2020 experimental paper in Physical Review C described irradiation of erbium and titanium deuterides with photons in the 2.5–2.9 MeV range. Deuterium provided densely packed fuel inside the metal. Photons broke apart some deuterons, releasing neutrons and protons; subsequent collisions transferred energy to other deuterons. NASA’s explanation highlights an important distinction: a metal specimen can initially be near room temperature while individual particles inside it acquire energies sufficient for nuclear interactions. The process doesn’t require the entire specimen to behave like a hot fusion plasma.

The researchers reported 2.45 MeV neutrons consistent with deuterium–deuterium fusion, as well as approximately 4–5 MeV neutrons that they interpreted as evidence of additional processes. Their analysis considered electron screening, in which surrounding negative charge modifies the electrical barrier encountered by positively charged nuclei. NASA’s public account describes the fuel as roughly a billion times denser than in magnetic-confinement tokamak plasmas. That concentration is useful, but the experiment still required an energetic trigger; the published findings weren’t a demonstration of useful net power.

Nagel was careful about this distinction. He respects the NASA work but questions whether its high-energy-photon initiation involves the same physics as what he calls conventional LENR. A NASA-hosted presentation by Lawrence Forsley in August 2026 places these experiments within a broader history encompassing plasma loading, co-deposition, electron screening, and Google-supported research. Forsley connects those developments to ARPA-E and DARPA’s subsequent programs. It’s a useful institutional history, but it shouldn’t be mistaken for a finding that every approach shares one mechanism.

ARPA-E Funds LENR Materials and Measurement

The Department of Energy’s Advanced Research Projects Agency–Energy gave this research a defined federal funding structure in 2023, selecting eight projects for approximately $10 million. The recipients included universities, a national laboratory, and smaller organizations. The objective was to determine whether LENR offered a promising scientific and technological direction—not to announce that the central questions had already been resolved. For a field long constrained by disputes over apparatus, materials, and interpretation, funding the means to test claims was itself consequential.

Texas Tech University’s $1.15 million award, Advanced Materials Characterization and Nuclear Product Detection for LENR, closely matches the problems Nagel described. The project focused on material fabrication, characterization, analysis, and advanced detection of nuclear products, including support for other teams in the initiative. ARPA-E lists its award period as June 2023 through July 2026 and now classifies it as an alumni project. That record establishes the funding and objectives; it isn’t, by itself, a report that the team successfully demonstrated LENR.

The University of Michigan received separate awards of $902,213 for radiation-detection capabilities and $1,108,412 for evaluating reported excess heat from deuterated palladium–nickel nanocomposites. The latter proposed combining microwatt-resolution calorimetry, neutron and gamma detectors, mass spectrometry, and computation. That arrangement directly addresses the problem of relying on one ambiguous signal. Researchers could examine whether thermal behavior coincided with nuclear or chemical products, rather than asking a heat measurement alone to establish the nature of an unfamiliar process.

Lawrence Berkeley National Laboratory’s $1.5 million project proposed measurements in metal hydrides at external excitation energies below 500 electronvolts. MIT received $2 million for work involving gas-loaded metal–hydrogen systems and nuclear diagnostics. Stanford investigated deuterated nanoparticles, while Energetics Technology Center and Amphionic pursued co-deposition detection and controlled nanoscale structures. Together, the projects formed a portfolio of competing approaches and measurement methods. That’s an important difference from funding a single reactor design: the program was structured to investigate which experimental conditions deserved further attention.

DARPA Puts Numbers on the Challenge

DARPA’s program makes the engineering question particularly explicit. MARRS stands for Mechanisms for Amplification of fusion Reaction Rates in Solids. Run through the Defense Sciences Office, it seeks quantitative explanations for reaction-rate enhancement and ways to increase those rates under controlled conditions. Its stated temperature range extends below roughly 2,000 kelvin, so “low temperature” here is relative to fusion-plasma conditions, not necessarily room temperature. DARPA’s own description acknowledges that absolute reaction rates remain low even where reported enhancements over earlier models are large.

The January 2026 program presentation sets targets against a stated baseline of approximately 0.1 reaction per second per gram. Phase one aims for 1,000 reactions per second per gram; phase two targets one million. The proposed structure has two 18-month phases, with predictive modeling accompanying experimental progress. Those are research goals, not announced achievements. Their value is that they provide measurable milestones against which teams can test whether particular interventions actually amplify nuclear reactions.

DARPA organizes those interventions into three areas: host materials, fuel loading, and reaction triggers. Materials questions include electron screening, hydrides, alloys, and vacancies. Loading includes electrochemical and gas-phase methods, with fuel mobility explicitly identified. Triggers include electronic and lattice excitations. The resemblance to Nagel’s account is striking: composition, structure, concentration, and movement are being treated as adjustable experimental variables. The agency is asking researchers to quantify how those variables interact, rather than merely identify an unusual specimen.

The rate targets also need a sense of scale. Using the presentation’s approximate energy of 3.5 MeV per reaction, one million reactions per second per gram converts to about 0.56 microwatts per gram before losses. That’s an illustrative calculation, not a forecast of a complete reactor’s output. It explains why a seven-order-of-magnitude improvement could be a meaningful scientific milestone without immediately producing a practical generator. Establishing reproducible control over a weak effect and engineering useful net power are different stages of development.

A New Generation of Cold Fusion Experiments

Research published outside our interview helps explain the renewed interest. A 2025 Nature paper, Electrochemical loading enhances deuterium fusion rates in a metal target, described a 300-micrometer palladium target separating an electrochemical cell from a vacuum chamber. A 30-kilovolt accelerating potential drove deuterium ions into the metal, while a separate 200-milliampere electrochemical current supplied additional deuterium. The arrangement allowed the researchers to change fuel loading while continuing the externally driven fusion experiment.

The team reported an average fusion-rate increase of 15%, with a quoted uncertainty of two percentage points, when electrochemical loading supplemented ion bombardment. The authors also reported a neutron-yield energy equivalent of only about one nanowatt, compared with a quoted 15-watt input. The result was therefore a measurement of how electrochemistry influenced a beam-driven nuclear process, not a net-energy demonstration. Its significance lay in the ability to manipulate one part of the system and observe a corresponding nuclear response.

A July 2026 Nature Communications paper extended the investigation into lower energies. Its authors combined electrochemical loading with ion bombardment in palladium and titanium, examining center-of-mass energies from approximately 0.25 to 6.5 keV. They reported a fusion-yield plateau below about 2 keV rather than the expected steep decline. At the lowest energies, the claimed enhancement exceeded 10¹⁸ relative to bare-nucleus calculations. That enormous ratio compares the observation with an extremely small theoretical expectation; it isn’t an energy-gain factor.

The 2026 authors also reported no excess fusion-like particle signals during electrochemical loading without the ion beam. That boundary matters. These experiments don’t automatically reproduce the original Fleischmann–Pons excess-heat claims. They address a narrower question: how strongly can a solid’s electronic structure, hydrogen content, and local environment influence a nuclear reaction initiated by an external beam? The opportunity is to make those influences measurable and controllable, then determine how far they can be extended. That’s a more precise research objective than treating “cold fusion” as one undifferentiated claim.

The Hidden Variable May Be the Material

Nagel kept returning to the same point: “Materials are absolutely central.” In his view, the problem isn’t adequately specified by saying that an experiment used palladium, nickel, or an alloy. Real metals contain vacancies, where atoms are missing; dislocations, where the lattice is distorted along a line; grain boundaries; and three-dimensional openings such as voids and gaps. Different LENR researchers favor different candidate environments. Nagel sees broad interest in defects, but he doesn’t claim that the field has identified one universally accepted active structure or explained its nuclear behavior.

His simplest illustration is a dime-sized piece of titanium deuteride, TiD₂, containing two deuterium atoms for every titanium atom. It holds an extraordinary concentration of fuel, yet it doesn’t become a power source merely by sitting in the laboratory. Nagel uses it to separate loading from stimulation: density is one condition, while the interactions needed for a reaction are another. He also sees an opportunity to borrow from hydrogen-storage research, where investigators have spent decades studying which materials absorb hydrogen and how they retain or release it.

At George Washington University, Nagel and fellow materials researcher Ashraf Imam are pursuing compacted powders and flash sintering. As Nagel describes it, the goal is to join powder grains sufficiently to make specimens that can be handled and tested while preserving potentially useful intergranular structures. That creates a practical trade-off. Loose powders are difficult to work with and can pose handling hazards, but processing them into a solid could also change the gaps and interfaces under investigation. The fabrication method is therefore part of the scientific question, not simply preparation for it.

He also proposes a systematic series of experiments in which palladium is mechanically worked and then annealed at different temperatures for different durations. The objective would be to deliberately vary grain size, boundaries, and other structural features, then compare those changes with experimental performance. It’s the kind of program that requires skilled staff, characterization equipment, and many controlled samples. Nagel’s standard for theory is equally demanding: an idea must lead to equations, numerical predictions, and comparisons with experiments. Identifying a possible reaction site doesn’t, by itself, explain how the reaction occurs.

Why Fuel Flux Matters

The distinction between fuel inventory and fuel transport offers another experimentally accessible variable. Loading describes how much hydrogen or deuterium is present in the metal. Flux describes its movement. Nagel is particularly interested in electromigration, in which an electrical drive along a loaded conductor redistributes hydrogen isotopes. He argues that this can influence both local concentration and the frequency of interactions. It’s a hypothesis about how to create favorable conditions, not a claim that moving hydrogen through any metal will necessarily produce nuclear energy.

That interest leads him back to a 1996 report associated with Italian physicist Giuliano Preparata. During our conversation, Nagel recalled an electromigration experiment using a long, thin palladium wire and reporting excess heat on the scale of tens of watts. Dividing that output by the small cathode volume, he said, produced power-density figures exceeding those associated with fission fuel rods. He was recalling some dimensions from memory, so the comparison needs to be traced to the exact experiment rather than treated as a finished device specification.

The denominator is crucial. Power normalized to a small wire volume can look impressive even when the total output is modest, and it says little by itself about the volume of a complete generator. A modern evaluation would need accurate wire dimensions, complete accounting of electrical inputs, calibrated thermal measurements, and a record of how long the reported excess persisted. Measurements of possible nuclear products would strengthen the investigation. The meaningful question is whether the same material and operating conditions reliably produce the same result—not whether a striking number can be calculated from one run.

Nagel says his GWU laboratory has the setup to revisit the experiment, but he’s short of help. He hopes to repeat it in the new year and encourages other researchers to do the same. That’s a revealing contrast with the scale of the federal programs: an experienced investigator can have a specific apparatus and a sharply defined question, yet staffing still determines whether the experiment happens. For Nagel, revisiting an old claim with modern instruments isn’t a retreat into history. It’s unfinished experimental work that could either strengthen the field or clarify where its expectations need to change.

The Test Is Whether Someone Else Can Make It Work

The military and space applications explain why these questions attract institutional attention. During our conversation, we discussed remote installations, maritime equipment, and power systems that could reduce dependence on difficult fuel deliveries. Nagel pointed to the expense of supplying energy to forward bases. These are prospective applications, not deployments, and their requirements differ substantially. A watt could matter for one device, ten watts for another, while household-scale uses would demand kilowatts. A successful technology would have to meet a particular application’s needs for output, mass, operating life, control, and maintenance.

Durability could be as important as peak performance. Nagel worries that the microscopic structures implicated in LENR might change under prolonged operation. At temperatures of a few hundred degrees Celsius—he discussed approximately 200–500°C—diffusion can alter defects and microstructure. If particular structures are responsible for activity, their evolution could reduce performance. He and colleagues are therefore considering whether active materials might eventually need to be renewed inside a generator. That’s a proposed engineering response to an anticipated problem, not an established maintenance procedure for a validated LENR product.

His development sequence is straightforward: first a laboratory experiment, then a prototype that somebody else can operate, and finally a product that can be manufactured, distributed, maintained, and recycled. The institutional programs address different portions of that progression. The Navy’s work exposed difficult replication questions; NASA examined specific ways to initiate reactions in deuterated solids; ARPA-E funded materials and diagnostics; DARPA established reaction-rate and modeling targets. Together, they document serious investigation. They don’t remove the requirement for independently reproducible performance or complete system-level energy accounting.

Near the end of our conversation, I suggested that a successful technology might eventually help power large AI data centers. Nagel immediately qualified the timetable: he wasn’t promising that outcome in two years. “This is a slog. It has been and will be,” he said. What he wants is sustained, systematic work of the kind that transformed semiconductor manufacturing—controlled processing, detailed measurements, repeated testing, and relentless improvement. The most important next milestone isn’t another impressive institutional name attached to the subject. It’s an experiment, and eventually a device, that someone else can reliably make work.

References

Primary interview and background

U.S. Navy research

NASA research

ARPA-E research programs

DARPA’s MARRS program

Recent experiments and technical background