Solid State Fusion & Diagnostics Engineering

What did two chemists claim in 1989, and why won’t it go away?

In 1989, two electrochemists, Martin Fleischmann and Stanley Pons, reported something that should not have been possible.¹ They ran electricity through heavy water, in which ordinary hydrogen is replaced by deuterium (a heavier form of the same atom), using an electrode made of palladium. Palladium has a useful trick: it soaks up hydrogen and its heavier cousin the way a sponge soaks up water, packing the atoms into the gaps in its crystal until there is nearly one deuterium atom for every palladium atom. Their cell, they said, gave off more heat than any chemical reaction they could name could account for. The explanation they reached for was nuclear fusion, at room temperature, in a beaker.

That claim collided with almost everything known about how fusion works, and most attempts to repeat it failed. It should have died there. It did not, because a stubborn handful of results kept turning up that no one could cleanly explain away. The field that studies them now goes by several names, among them solid-state fusion and condensed matter nuclear science.

If it really is fusion, where are the neutrons?

The objection that has done the most damage to the claim is worth understanding, because it is not a matter of taste. When two deuterium nuclei fuse, nuclear physics says they almost always split one of two ways. About half the time you get a tritium nucleus and a proton. The other half you get a helium-3 nucleus and a free neutron. Only about once in ten million fusions do you instead get helium-4 and no neutron at all.

A fusion reaction fast enough to warm a cup of water should therefore spray out neutrons in enormous numbers. You could stand across the room with a neutron counter and see them. The cold fusion results report close to the opposite: heat, sometimes a trace of helium-4, and little or no matching neutron flux. Standard nuclear physics has no room for that pattern. Either something is happening that no one understands, or the measurements are wrong. That gap, more than anything else, is why the odds of a genuine nuclear effect sit well below even. Thirty years of failed replication push the same way, including a 2019 multi-institution effort that set out to retest the original claims to a modern standard and found no excess heat.²

Notice what kind of objection this is. It is an argument about counts: how many neutrons, at what energy, measured against what background. It is answerable with instruments.

Why the whole fight comes down to three instruments

Strip the field down to what it actually asserts and you are left with three measurements.

The first is heat. A calorimeter measures how much energy leaves the cell, and the claim is that more comes out than the electricity puts in. The second is helium. A mass spectrometer, which sorts atoms by weight, is supposed to find helium-4 in a cell that has no chemical business making any, the “ash” that would tie the heat to a nuclear source. The third is particles: a detector that registers the neutrons or charged particles a nuclear reaction ought to throw off.

Every argument the field has had, in either direction, comes back to whether those three instruments were read correctly. That is unusual. In most of physics the measurement is routine and people argue about what it means. Here the meaning is genuinely open, and the measurements themselves are where the fight lives. This is useful to know, because it tells a would-be investigator where to stand. You do not need a new theory of the nucleus to make progress. You need better instruments and the discipline to read them honestly. That discipline has a name: diagnostics engineering.

Reducing a thirty-year controversy to three measurements is a simplification. It is also the lens that shows where the disagreements are actually decidable.

Can you trust a calorimeter reading a few hundred milliwatts?

The heat signal is small, and that is the whole difficulty. A cell might draw a watt or more of electrical power and return only a few hundred milliwatts of claimed excess. You are looking for a small difference between two large numbers, one of the most treacherous things an experimentalist can be asked to do. Every slow drift and every imperfect assumption about where the heat is going lands right on top of the quantity you are trying to measure.

The sharpest published challenge to the excess-heat results is exactly this kind of argument. A chemist named Kirk Shanahan proposed that a small, unnoticed shift in a cell’s calibration could manufacture an apparent excess of about the right size, driven by a change in where inside the cell the heat is produced as the run goes on.³ A group of ten researchers replied that the calibration data do not support such a shift, and that it cannot survive cell designs which measure heat outside the cell entirely.⁴ You do not have to decide who is right to see the shape of the thing. Both sides are doing diagnostics engineering, arguing about calibration and systematic error, and the dispute will close on better calorimeters rather than on argument.

If you wanted to settle it yourself, you would build a sealed cell that recombines its own gases and measures heat leaving an outer jacket, run identical blank cells with ordinary water that should show nothing, and calibrate with a known electrical heater before and after every run. A signal that survives all of that is hard to wave away. One that vanishes tells you something too.

Helium-4: nuclear ash, or air leaking in?

Suppose the heat is real. The way to tie it to a nuclear origin is to find helium-4, the atom that deuterium fusing into helium would leave behind. Detecting it pushes mass spectrometry close to its limit, for two reasons.

The first is a coincidence of weight. A molecule of deuterium gas weighs almost exactly as much as a single helium-4 atom. The two differ by about 0.0256 atomic mass units, so an instrument that cannot resolve that sliver of a gap will report deuterium as helium and never know the difference. The second reason is that helium is already in the air around us, at roughly five parts per million, and it seeps into almost any sampling line through the smallest leak. The signal being claimed sits down at the level of parts per billion, a thousand times smaller than the background trying to get in.

So the reported link between excess heat and helium, at roughly the ratio you would expect if deuterium were turning into helium, rests on limited data and has not been widely reproduced with high-resolution instruments. Its two weak points are precisely leak-tightness and background subtraction. The fixes are what any careful metrologist would prescribe: instruments sharp enough to separate deuterium gas from helium-4, gas lines built entirely of metal with metal-sealed valves to keep air out, and a second gas, neon, watched as a tracer so that any air sneaking in announces itself. The place to push is the measurement.

A track in plastic: an alpha particle, or an artifact?

The third instrument gives the cleanest example of a distinction that runs through the entire subject. Certain plastics, one called CR-39 in particular, record the passage of a charged particle as a microscopic pit that can later be etched and counted under a microscope. This is not fringe equipment; the same plastic is a standard particle detector in mainstream fusion research. In some cold fusion cells, researchers have found reproducible populations of these pits, clustered right at the electrode where a reaction ought to be.

The pits are real and they repeat. What they mean is a separate question. Analyses of the pit sizes have argued that the dominant tracks cannot have come from the energetic alpha particles a fusion reaction would produce. So you can have an anomaly that is genuinely reproducible and still have its nuclear interpretation wide open. Holding those two facts apart, refusing to let “reproducible” quietly slide into “nuclear,” is the habit of mind the whole field demands.

What would it actually take to settle it?

The strongest evidence in a contested field is agreement across instruments that fail in different ways. Two unrelated calorimeter designs converging on the same excess, helium climbing in step with that heat, a loading gauge and a calorimeter telling one consistent story: no single instrument carries that weight, but together they can. Designing an experiment so its measurements are genuinely independent, and so that a null result is as informative as a positive one, is as much a contribution to how we know things as to what we know.

None of this asks you to believe in cold fusion. The honest starting point is skeptical. The conflict with well-tested nuclear physics, and more than thirty years of failed replication, put the odds of a real nuclear effect low.⁵ That is exactly why the measurement bar sits so high. And even a clean, repeatable link between heat and helium would establish only that something real is happening. Whether it is nuclear would still demand a mechanism, and a pattern of replication strong enough to move a well-supported physical prior.

Here is the thread a curious student can pull. Take one of the three measurements and ask what it would take to make it undeniable, in either direction. What calorimeter design removes the calibration ambiguity for good? What resolving power splits deuterium gas from helium-4 beyond dispute? What would turn a contested pit into a settled track? The field has a thirty-year backlog of measurements that were never quite good enough to answer even the first question, whether the phenomenon is real at all. Making them good enough is not a matter of belief. It is a job, and the tools for it already exist.

Notes

1. Martin Fleischmann and Stanley Pons, “Electrochemically Induced Nuclear Fusion of Deuterium,” Journal of Electroanalytical Chemistry and Interfacial Electrochemistry 261, no. 2A (1989): 301–308. ↩
2. Curtis P. Berlinguette et al., “Revisiting the Cold Case of Cold Fusion,” Nature 570 (2019): 45–51. ↩
3. Kirk L. Shanahan, “A Systematic Error in Mass Flow Calorimetry Demonstrated,” Thermochimica Acta 382 (2002): 95. See also his follow-up in Thermochimica Acta 428 (2005): 207. ↩
4. Jan Marwan et al., “A New Look at Low-Energy Nuclear Reaction (LENR) Research: A Response to Shanahan,” Journal of Environmental Monitoring 12, no. 9 (2010): 1765–1770. ↩
5. U.S. Department of Energy, Report of the Review of Low Energy Nuclear Reactions (Washington, DC: U.S. Department of Energy, Office of Science, 2004). ↩

Introduction: Three Measurements

Strip solid-state fusion down to what it actually asserts and you are left with three measurements. A calorimeter reads out more energy than went in. A mass spectrometer finds helium-4 where a room-temperature electrochemical cell has no business making any. A track detector or a neutron counter registers particles that ordinary chemistry cannot account for. Every argument the field has had for thirty-five years, in either direction, reduces to whether those instruments were read correctly. In the first instance that is not a physics question. It is a diagnostics question.

The point is worth stating plainly because it inverts the usual relationship between measurement and interpretation. In most of physics the measurement is routine and the meaning is where people fight. In SSF the meaning is genuinely open, but the measurements themselves are where the fight has always been. Excess power at the level of a few hundred milliwatts against a watt or more of input; helium at parts per billion against an ambient background near 5.2 ppm; charged-particle tracks argued over micrometer by micrometer. These are among the hardest measurements an experimentalist can be asked to defend, and they sit at the exact center of the field's credibility. Reducing the field to three measurements is a lens rather than a theorem, but it is the lens that shows where the disputes are actually decidable. Diagnostics engineering is therefore not a service function bolted onto SSF. It is the discipline that owns the deciding instruments.

I — Background: Where the Two Fields Actually Meet

Solid-state fusion, also called condensed matter nuclear science, investigates whether hydrogen isotopes loaded into a metal lattice at near-unity atomic ratios can produce nuclear-scale energy release. The originating claim was calorimetric: Fleischmann and Pons reported heat from palladium cathodes in heavy water beyond any chemistry they could identify. The reasons for mainstream skepticism are also, at bottom, about measured quantities. If deuterons were fusing fast enough to warm a calorimeter, the reaction should flood the apparatus with neutrons, because the d+d reaction goes to the tritium-plus-proton and helium-3-plus-neutron channels roughly half the time each, and to helium-4 only on the order of once in ten million events. The reported pattern is close to the opposite: heat and helium-4, little or no commensurate neutron flux. That deficit is the strongest single objection in the field, and it is an objection about counts and cross-sections, not about metaphysics.

A diagnostics engineer should read that standoff as a target-rich environment. The question that decides everything is whether a set of demanding measurements are real or artifactual, and that is a question the field's originating communities, electrochemists and nuclear theorists, are not always best equipped to close. What follows maps the intersection concretely: first the measurement problems where diagnostics engineering can move the field, then what the field offers back.

II — Ten Ways Diagnostics Engineering Advances SSF Research

1. Calorimetry and the systematics that decide it

Excess heat is the primary observable, and how it is measured is not a detail. Three calorimeter architectures dominate the literature: isoperibolic (constant surrounding temperature, heat inferred from a temperature difference and a calibration constant), mass-flow (heat carried away by a coolant of known heat capacity), and Seebeck-envelope (thermopile integration of all heat leaving a jacket). Each has a different dominant systematic. The most substantive published challenge to the whole excess-heat corpus is a calorimetric one: Shanahan argued that an unrecognized shift in the calibration constant, driven by a change in where heat is generated inside an open cell as recombination moves to the electrode, can manufacture an apparent excess of the right size. A ten-author group replied that the shift is not supported by the calibration data and cannot survive a switch to closed-cell, mass-flow, or Seebeck designs that measure heat outside the cell. Whichever side is right, the exchange is a diagnostics-engineering argument conducted in diagnostics-engineering terms, and it will be settled by better calorimeters, not by theory.

2. Calibration, blanks, and the discipline of the null

The calibration-constant debate points at a general contribution: rigorous reference protocols. In-situ Joule heaters that inject a known power, blank cells run with light water or platinum cathodes, and calibrations repeated across the full current range before and after an active run are what separate a defensible excess from a drifting baseline. A field whose signal is a small difference between two large numbers lives or dies on its uncertainty budget. Bringing traceable calibration and a documented error model to every reported watt is among the most useful things an outside metrologist can do here, and it is unglamorous work that the field has not always done well.

3. In-operando loading metrology

The one experimental variable most consistently linked to excess heat is the loading ratio, the atomic fraction of deuterium to palladium. It is measured in real time from the electrical resistance of the cathode, normalized to its unloaded value: the R/R₀ method, which exploits the known relationship between interstitial occupancy and resistivity and has been the field's standard loading gauge for three decades. Reported excess heat clusters above roughly D/Pd ≈ 0.85–0.90, though high loading appears necessary rather than sufficient: a 2019 multi-institution collaboration that set out to retest the cold-fusion claims to a high modern standard reported no excess heat across its sample set, while stressing that reaching and verifying high loading is itself difficult, and read the extreme-loading regime as underexplored rather than closed. A better in-operando gauge of loading, and of its spatial uniformity along a cathode, is a direct diagnostics contribution to the field's most reproducible correlation.

4. Helium-4 detection at the interference limit

Helium-4 is the ash that would tie heat to a nuclear origin, and detecting it is a mass-spectrometry problem near the edge of the technique. The obstacle is a mass interference: singly ionized D₂ sits at m/z 4 essentially on top of ⁴He⁺, separated by about 0.0256 u, so distinguishing them demands a resolving power beyond a routine quadrupole. Against that sits the ambient helium background of roughly 5.2 ppm, which leaks and permeates into almost any sampling train. The field's response has been exactly what a diagnostics engineer would prescribe: high-resolution sector instruments, all-metal gas-handling with metal-gasketed valves to defeat atmospheric in-diffusion, and ²⁰Ne monitoring as an air-contamination tracer. The reported correlation between excess power and helium, at an energy ratio in the neighborhood of the d→⁴He value, has been reported by more than one group but rests on limited datasets, with disputed effect sizes and little high-resolution independent replication; it is presented here as reported, and its main vulnerabilities are precisely leak integrity and background subtraction. That is why the measurement, not the claim, is the place to push.

5. Charged-particle and neutron diagnostics

Detecting the particles a nuclear process should emit, and bounding the ones it apparently does not, is the second radiation-diagnostics front. CR-39 solid-state track detectors have recorded reproducible pit populations in Pd/D co-deposition cells, coincident with the cathode. The pits are real and repeatable; their interpretation is genuinely disputed, with pit-diameter analyses arguing that the dominant tracks cannot come from MeV alpha particles. This is the cleanest illustration in the whole field of the distinction diagnostics work must protect: an anomaly can be reproducible and its assignment to a nuclear origin still be open. CR-39 is not fringe hardware; it is a calibrated charged-particle spectrometer in mainstream inertial-confinement fusion. On the neutron side, He-3 and BF₃ counters and bubble detectors must establish upper limits on a flux that is at most sporadic and low as reported, which is a demanding background-and-coincidence problem in its own right, and the resulting limits are what give the missing-neutron objection its quantitative teeth.

6. Platform engineering for reproducibility

Reproducibility in SSF is partly a hardware problem. Closed cells that recombine evolved gas internally remove the largest calorimetric ambiguity of open cells. Controlled thermal environments, defined electrode preparation, and instrumented control cells run in parallel turn a one-off observation into a comparison. Much of the field's irreproducibility is plausibly traceable to uncontrolled variables in materials and cell construction rather than to the absence of an effect, which is a design brief an engineer can act on directly.

7. Safety engineering, with a real precedent

These experiments combine deuterium and hydrogen gas, high electrolysis currents, and stored chemical energy in the oxygen and deuterium evolved during electrolysis. The hazard is not hypothetical: a 1992 explosion in an SRI cold-fusion cell killed a researcher and injured three others. A forensic analysis by Lawrence Livermore attributed the blast to a chemical gas event in the cell rather than to any nuclear or anomalous-heat process, finding no tritium or neutron-activation signatures. That is exactly why it belongs here: ordinary deuterium–oxygen chemistry at these currents is dangerous on its own terms. Hazard analysis, containment and venting design, and interlocked shutdown are load-bearing engineering here, not compliance paperwork, and they are prerequisites for the kind of long-duration unattended runs the calorimetry needs.

8. Signal processing and low-SNR extraction

SSF data are noisy, intermittent, and gathered over long baselines, which is the natural habitat of signal processing. Modeling the calorimeter as a thermal transfer function, propagating uncertainty through a full power balance, and testing a heat-helium or heat-loading correlation with methods that respect autocorrelation and multiple comparisons are the difference between a suggestive plot and a defensible result. The temporal correlation between excess power and helium accumulation, if it holds up, is exactly the kind of claim that stands or falls on disciplined time-series analysis rather than on any single data point.

9. Instrumentation for scale-up, conditionally

If a reproducible excess-heat effect were established, the engineering questions would shift to power measurement at higher coefficient of performance, calorimetry that stays accurate at useful power densities, and in-situ diagnostics embedded in a prototype rather than a benchtop cell. This is stated in the conditional deliberately. It is a reason for an instrumentation engineer to find the problem interesting, not a forecast that the condition will be met.

10. Cross-method corroboration

The strongest evidence in a contested field is agreement across independent methods with independent failure modes. Heat measured by mass-flow and by Seebeck calorimetry converging, helium rising in step with power, a loading gauge and a calorimeter telling a consistent story: these carry weight that no single instrument can. Designing an experiment so that its diagnostics are genuinely orthogonal, and so that a null is as informative as a positive, is a diagnostics-engineering contribution to epistemology as much as to hardware.

III — Five Ways SSF Research Advances Diagnostics Engineering

1. Sub-watt calorimetry over long baselines

Resolving a few hundred milliwatts of steady excess against a large input, stable for days, pushes calorimetric metrology into a corner it rarely has to occupy. The demands on baseline stability, calibration traceability, and drift rejection are transferable well beyond SSF, to any application that needs to trust a small persistent thermal signal.

2. Ultra-trace noble-gas detection against a leaky background

The helium problem, quantifying a parts-per-billion signal at m/z 4 against an ambient background that permeates the apparatus, is a hard case in trace mass spectrometry and leak-tight sampling. The methods developed to defeat it, all-metal trains and isotopic air tracers, carry over to isotope geochemistry and to tritium and noble-gas handling facilities.

3. Automated track analysis

The dispute over CR-39 pit interpretation is, in effect, a call for objective, high-throughput track classification. Automated pit-parameter analysis and machine-learning classification of track morphology are under active development for exactly this reason, and they feed back into inertial-confinement-fusion diagnostics and radiation dosimetry, where CR-39 is already standard.

4. Control and blinding discipline for irreproducible signals

A field where a real effect might appear in one run of ten forces the measurement community to take blinding, pre-registered protocols, parallel control cells, and open data seriously. The metascience of detecting a rare, intermittent, weakly reproducible signal is a general methodological problem, and SSF is an unusually sharp test case for it.

5. Instrumenting extreme material states

Palladium loaded past D/Pd ≈ 0.9, co-deposited films, and nanostructured hydrides are unusual condensed-matter systems in their own right, independent of any nuclear question. Characterizing them in operando under extreme hydrogen loading, mechanically, electrically, and structurally, is a legitimate materials-diagnostics challenge, and clean mainstream measurements on related loaded-metal systems show the payoff of doing it well.

IV — Strategic Note

A diagnostics engineer may not think of SSF as their problem, and the field's history with contested nuclear claims has made many technical communities keep their distance. But the actual work is squarely in scope. The field's load-bearing assertions are measurement assertions, and measurement is what diagnostics engineering does. The contribution needed is not endorsement of any mechanism. It is better calorimeters, cleaner helium analyses, defensible particle counts, and honest uncertainty budgets.

This cuts both ways, and that is its appeal. The same rigor that could put a reproducible effect beyond reasonable doubt could also expose an artifact beyond reasonable doubt. Shanahan's calibration-shift argument and the closed-cell rebuttal are both diagnostics work; the 2004 Department of Energy review split precisely on whether the measurements were sound, and identified near-background helium and calorimetric ambiguity as the cruxes. An engineer who makes the signal or the artifact undeniable has done the field a service in either case. The frontier is not a belief to be adopted. It is a set of measurements waiting to be made properly.

V — Conclusion

The distance between solid-state fusion and diagnostics engineering is smaller than the field's reputation suggests, because the questions that matter here are measurement questions. Whether a calorimeter is telling the truth, whether helium at parts per billion is nuclear ash or a leak, whether a track came