Electrochemically-enhanced nuclear fusion of deuterium

Abstract

The ultimate goal of nuclear fusion research is to achieve a reaction that generates more energy than it consumes. Fuels such as deuterium need to be confined to increase the collision probability of particles. We therefore set out to investigate whether electrochemically loading a metal lattice with deuterium fuel could increase the probability of nuclear fusion events. Here, we report a benchtop fusion reactor that enabled us to bombard a palladium metal target with deuterium ions. These deuterium ions undergo deuterium-deuterium (D-D) fusion reactions within the palladium metal. We showed that the in situ electrochemical loading of deuterium into the palladium target resulted in a 15(2)% increase in D-D fusion rates. This experiment shows how electrochemical reactions at the eV energy scale can impact nuclear reactions at the MeV energy scale.

Main

Nuclear fusion research for energy applications aims to create conditions that produce more energy than required to initiate the process. Both magnetic and inertial confinement fusion are being widely investigated to reach this goal; however, they demand extreme environments, specialized infrastructure, and prolonged timelines for commercial deployment.

In addition to confinement time and temperature, the Lawson Criterion states that fusion reaction rates are also governed by the density of the fuel. A higher density of fuel (e.g., deuterium) increases the probability of particle collisions. Magnetic and inertial confinement fusion typically achieve fuel densities of approximately 10²⁰ m⁻³ and 10³¹ m⁻³, respectively, where the latter density exists for only hundreds of picoseconds. It is therefore notable that a deuterium fuel density of 10²⁸ m⁻³ can be easily achieved in a solid metal lattice (Supplementary Note 1).

In 1934, Oliphant and Rutherford conducted the first demonstration of deuterium-deuterium (D-D) nuclear fusion, where they bombarded a target of solid metal, plated with deuterated material, with high-energy deuterium ions. The incident deuterium ions collided with deuterium embedded in the target. Fusion reactions using liquid, gas, and plasma targets have since been demonstrated. For experiments that use solid metal targets, a high-energy deuterium ion (D⁺) from a beam — generated using a particle accelerator, a high-power laser, or using gamma rays — strikes the metal target and becomes embedded within the target. A subsequently delivered D⁺ from the beam can collide with the embedded D⁺ to create a D-D fusion event. Fusion rates within deuterated metals have been observed to surpass rates with gaseous targets.

Building on these experimental observations, we set out to independently load a metal with high concentrations of deuterium fuel to increase the probability of D-D fusion events with a metal target. To achieve this objective, we designed and built a benchtop particle accelerator, the "Thunderbird Reactor", to: (i) deliver D⁺ into a palladium target (by plasma immersion ion implantation; "PIII") to mediate D-D fusion events within a palladium metal lattice; and (ii) electrochemically load a palladium target with deuterium sourced from D₂O(l) in an adjacent aqueous electrochemical cell. The reactor is capable of detecting and quantifying neutrons formed during the D-D fusion events that occur within a rigid palladium metal foil that physically separates a plasma reactor under vacuum from an aqueous electrochemical cell.

The Thunderbird Reactor delivers D⁺ sourced from a "plasma thruster" (vide infra) through a vacuum chamber towards the palladium target. A dedicated power supply with an applied voltage ("sheath voltage") selectively accelerates the D⁺ into the target. As the beam of incident D⁺ loads deuterium into the palladium target ("Phase I"), the rate of neutron production increases for a period of time, before reaching a constant rate of neutron production. This experiment represents the first demonstration of PIII-induced D-D fusion. We also demonstrate that while maintaining an incident beam of D⁺ on the palladium, electrochemically loading the palladium with deuterium from an adjacent electrochemical cell ("Phase II") further enhances D-D fusion events by more than 10%. This experiment provides direct evidence that the electrochemical loading of a metal target at the eV energy scale can increase nuclear fusion rates at the MeV energy scale.

Fig. 1: The Thunderbird Reactor. a, Working principle of the Thunderbird Reactor. Deuterium gas (D₂) is fed to the plasma thruster through the deuterium gas inlet. The D₂ is ionized by a microwave source. Plasma thruster magnets expel the resultant plasma of D⁺ and e⁻ into the vacuum chamber. A minus 30 kV voltage applied across the palladium target and the vacuum chamber creates a plasma sheath enriched in D⁺, and accelerates the D⁺ into the palladium target where nuclear fusion can occur. On the opposite side of the palladium target is an electrochemical cell that oxidizes OD⁻ into D₂O and O₂(g) at the anode. The D₂O is then reduced to D at the palladium target, which acts as a cathode. The D is then absorbed into the palladium target to potentially fuse with D sourced from the plasma thruster. The plasma current is set at 0.5 mA by the power supply and a galvanostat maintains a constant current of 200 mA across the anode and cathode of the electrochemical cell.

The Thunderbird Reactor

The Thunderbird Reactor (Fig. 1) is a bespoke bench top sized particle accelerator designed and built to achieve electrochemically enhanced D-D nuclear fusion. The three main components of the reactor include a plasma thruster, a vacuum chamber, and an electrochemical cell. We designed the reactor to be capable of plasma ion implantation to drive D⁺ into a metal target to produce fusion reactions within a metal lattice. We selected a 300-µm thick palladium foil because palladium is capable of hosting high concentrations of deuterium. The D⁺ were generated in a custom-made plasma thruster driven by a 2.45-GHz microwave generator, then passed into a vacuum chamber by a set of axial ring magnets. Within the vacuum chamber set at 10⁻⁵ Torr, the mean free path of electrons and D⁺ greatly exceed the dimensions of the chamber (Supplementary Note 2, Supplementary Fig. 1). The plasma jet can therefore travel unobstructed from the plasma thruster to the palladium target. A high negative voltage applied to the palladium target repelled electrons and created a "plasma sheath" close to the target. This plasma sheath is enriched in D⁺, and is visible to the eye (Fig. 1). A power supply created a sheath voltage to accelerate the D⁺ entering the sheath to keV energies and implant them in the target (Fig. 1). Note that this design does not require complex ion optics and accelerating grids, thus the footprint of the Thunderbird Reactor is merely 120×80×70 cm³, and fits on a standard laboratory bench. The reactor is operated using a custom built software and interface (Supplementary Note 3).

On the opposite side of the palladium target, an electrochemical cell contained 2 M K₂CO₃(aq) in D₂O. The D₂O was reduced at the palladium target into D and OD⁻ (Eq. 1). The oxidation of OD⁻ at an iridium anode formed D₂O and O₂(g) (Eq. 2). The D atoms were absorbed into the palladium lattice more readily than D-D recombination to form D₂(g). The electrochemical cell was inspired by our previous work on palladium membrane reactors and modified to function within the vacuum chamber (Supplementary Note 4, Supplementary Fig. 2). Note that D₂(g) could form on either face of the palladium during electrolysis. The palladium therefore served as a cathode and membrane for the electrochemical cell, a target for D⁺ sourced from the plasma thruster, and a physical separator between the vacuum and the electrochemical cells.

Cathode reaction: D₂O + e⁻ → D + OD⁻ (Eq. 1)

Anode reaction: OD⁻ → ½ D₂O + ¼ O₂ + e⁻ (Eq. 2)

The fusion of two deuterium atoms is expected to produce, with equal probability, (i) a neutron (n) and helium-3 (³He) with kinetic energies of 2.45 MeV and 0.82 MeV, respectively (Eq. 3); or (ii) a proton (¹H) and tritium (³H) with kinetic energies of 3.02 MeV and 1.01 MeV, respectively (Eq. 4). A third possible channel is the production of ⁴He and a γ-ray with an energy of 23.8 MeV (Eq. 5), but there is a low probability (10⁻⁵ %) of this reaction occurring and is not considered here.

D + D → ³He + n⁰ (Eq. 3)

D + D → ³H + ¹H (Eq. 4)

D + D → ⁴He + γ (Eq. 5)

To test whether nuclear fusion rates could be enhanced upon electrochemical loading of the palladium target, we measured neutrons as a proxy for fusion events. Because the walls of the vacuum chamber are effectively transparent to 2.45 MeV neutrons, we used a neutron sensitive scintillation detector coupled with a photomultiplier tube, positioned 12 cm from the palladium target. This detector was located outside of the vacuum chamber for protection from the harsh environment within the reactor. The detection system enabled the differentiation of neutrons from background gamma rays using pulse-shape discrimination of the digitized detector signals (Fig. 2, Supplementary Note 5, Supplementary Fig. 3 and 4). Consequently, we were able to exclude more than 99.9999% of gamma rays.

Nuclear fusion, by plasma immersion ion implantation

To validate fusion events within the Thunderbird Reactor, we first performed a control experiment within each experimental campaign. Before turning on the reactor for any experiment, we would collect 5 minutes of background neutron measurements. This neutron production rate was measured to be a rate of 0.2 neutrons per second (n/s) in our laboratory environment. For all of our experiments including the control experiment, we fed D₂(g) into the thruster at a rate of 0.5 sccm and turned on the plasma thruster by using a 200-W burst of microwave power over 5 seconds to initiate the ionization process. We then applied 20 W to sustain operation. The plasma current was held constant at 0.5 mA to control for neutron yield. The target became engulfed in plasma (Fig. 1). The high voltage (minus 30 kV) across the target and vacuum chamber, defined herein as "sheath voltage", would then repel negative charge from the palladium to create a plasma sheath. The D⁺ within the plasma sheath were accelerated to an energy of 30 keV to strike the target and induce D-D fusion.

After activating the plasma thruster and applying the sheath voltage, the neutron production rate rose from the background value of 0.2 n/s to a stable value of 130 to 140 n/s after approximately 30 minutes (Fig. 3). We defined the neutron production rate to be stable when the neutron production rate remained within ±5% of its mean value over a 30-minute period after the sheath voltage was applied. The measured neutron energy was consistent with D-D fusion reactions, when benchmarked against the experimental energy distribution of neutron data simulated with the Monte Carlo N-Particle transport code (MCNP 6.2) and the detector response function (DRiFT) toolkit (Supplementary Note 5, Supplementary Fig. 5 and 6).

The initial increase and eventual saturation of the neutron production rate during the beam-loading experiment is consistent with fusion occurring within the lattice of the metal target, and not within the gas or plasma phase (Supplementary Note 6, Supplementary Fig. 7).

Fig. 2: Neutron diagnostics and discrimination. a, The 3D distribution showing the neutron counting window (red). The red outline shows the particles that are identified as neutrons, separate from the gamma-ray distribution. b, A slice of the 3D distribution taken at a single 15 keVee slice, showing the gamma ray and neutron production rates for a light output of 195 keVee. The solid (black) lines are the mean of the gamma-ray (gray) and neutron (red) distributions, each with a standard deviation. The dashed (red) line delineates the gamma-ray window, placed at the mean plus five standard deviations; more than 99.9999% of all gamma-rays are contained in the region to the left of this line. Those particles found to the right of this line are classified as neutrons. c, The results of seven independent beam-loading experiments at minus 30 kV and 0.2 mA using different Pd targets showing that neutron production rates are approximately the same 70 n/s after 120 min of reactor operation.

Electrochemically-enhanced fusion

The next stage of our experiments was to test if the electrochemical loading of the palladium target would increase the neutron production rates. To test for this effect, we would, after the control experiment, remove the target from the reactor and then anneal the target at 400 °C for one hour at 10⁻⁵ Torr to remove deuterium from the metal. X-ray diffraction (XRD) characterization of the Pd target, before and after annealing, confirmed that the annealing successfully removed deuterium from the Pd lattice (Supplementary Fig. 8). We then inserted the Pd target back in the reactor, and repeated the same procedure described above for the control experiment. If the neutron production rate profiles matched the control experiment, then we would move on to the next stage of electrochemically loading the Pd target. After the reactor reached a stable neutron production rate (usually after 60 min), the adjacent electrochemical cell was turned on. A total current of 200 mA was applied across the anode and cathode, and was kept constant by using the power supply in galvanostatic mode. Deuterium levels in the vacuum chamber, which were detected by a residual gas analyzer, started to increase two minutes after turning on the electrochemical cell. This increase is consistent with deuterium being passed through the Pd target with the electrochemical cell. Independent in situ XRD characterization of the Pd target revealed that the electrochemical current successfully loaded the Pd target with a high concentration of deuterium (D/Pd molar ratio of 0.7), which remained in the Pd lattice under ambient conditions (Supplementary Figs. 9 to 13).

The most important aspect of this study is that we observed an increase in neutron production rate after the electrochemical cell was turned on. As shown in Fig. 3, the neutron production rates rose with time to reach saturation at 133.1(5), 140.3(5), and 136.0(3) n/s for three separate targets, representing an average 15(2)% increase in neutron production rate due to electrochemical loading.

If the electrochemical cell is turned off, the neutron production rates do not change by more than 2% of the maximum observed during the electrochemical cell "on" period (Supplementary Fig. 14). We independently confirmed using in situ XRD methods that the deuterium that is electrochemically loaded into the target does not exit the palladium membrane on a timescale relevant to the experiment under ambient condition or under vacuum (Supplementary Fig. 12 and 13). When the electrochemical cell was turned back on, the neutron production rate increased yet again, but to a lesser extent (6% increase) than the 20% increase in the initial cycle.

Fig. 3: Neutron production rates in the Thunderbird Reactor. Three separate experimental campaigns showing neutron production rate using plasma immersion ion implantation (PIII) as a function of time, with and without electrochemical loading of the palladium target. Each campaign was performed by first using PIII on the target for 2 hours (gray circles). The palladium target was then removed from the reactor, placed under vacuum at 400 ºC for 1 hour to remove the loaded deuterium, and then inserted again into the reactor. The same PIII procedure was repeated. After the rate of neutron production reached a constant value, the electrochemical cell was turned on to load palladium with additional D sourced from D₂O(l) (red squares). Note that Phase I (beam loading only) of the successive experiments produces similar neutron production rate profiles. The neutron production rates increase by 15.1(1)%, 11.3(1)% and 18.5(2)% with electrochemical loading (Phase II; red squares) for the three experimental campaigns shown here. The gray area before Phase I represents the collection of background neutron production rate prior to turning on the plasma thruster.

We built the Thunderbird Reactor to bridge chemical reactions and nuclear reactions that occur at the eV and MeV energy scales, respectively. Using the Thunderbird Reactor, we demonstrated that the electrochemically loading of a metal target can increase D-D fusion rates, driven by plasma immersion ion implantation, by an average of 15(2)%. While this experiment represents the first incontrovertible example of electrochemistry directly increasing nuclear fusion rates, the Thunderbird Reactor produces a neutron yield equivalent to merely 10⁻⁹ W with 15 W of input power (Supplementary Note 7). There are many ways the nuclear fusion, materials science and electrochemistry communities can increase this power output. For example, the deuterium fuel densities within the metal target can be increased by either using an inductively coupled plasma that can be maintained at higher pressures to make more ions available at the target, or by using metals capable of higher deuterium loadings (e.g., Nb, Ti). It is also intriguing to consider whether secondary fusion events in the target with ³He and tritium, both of which are products of D-D fusion, as fusion fuels. There is also an opportunity to leverage quantum coherence effects with target materials and with coherent stimuli. This reactor could also help answer fundamental questions related to screening effects and intranuclear resonances at lower energies (below 5 keV). The ability to influence the target within an accessible benchtop-scale fusion reactor, which is also coupled to an adjacent electrochemical cell capable of controlling local and transient fuel density, presents many new avenues of exploration for advancing the nuclear fusion sciences.

Methods

Materials

Potassium carbonate (K₂CO₃, 99%) was purchased from Thermo Fisher Scientific. K₂CO₃ was kept in a Napco 5831 oven at 120 °C for at least 24 hours to eliminate the residual water. Deuterium oxide (D₂O, 99.9%) was purchased from Cambridge Isotope Laboratories and used as received. Iridium wire (0.5 mm, 99.95%) was obtained from Taobao. Ag/AgCl reference electrodes (CHI111) were purchased from CH Instruments. Potassium chloride saturated in H₂O was purchased from Fisher Chemical and used as received. Potassium hydroxide (KOH, at least 85%, pellets) was purchased from Sigma Aldrich and used as a ⁴⁰K radioactive gamma-ray emitting source during the neutron detector calibration process.

Plasma thruster

The plasma thruster is custom-made from SS-316 stainless steel and includes three ring magnets and a set of hexapole magnets. A type N feedthrough, located below the plasma thruster, functions as an antenna to transmit microwaves into the plasma thruster (Supplementary Fig. 15). The type N feedthrough transmits the microwaves via a coaxial cable and is controlled through a proportional-integral-derivative (PID) loop. The deuterium flow rate is adjusted by a mass flow controller and fed into the gas inlet on the side of the thruster through SS-304 stainless steel pipes. The detailed components of the plasma thruster, along with their make and model, are as follows:

• Microwave antenna (KJL, SS, IFTNG012033M, Type N Feedthrough)
• Ring magnet (CMS Magnetics, NR011-42N, N42 neodymium)
• Bar magnet (20×10×10 mm N35 neodymium)
• Mass flow controller (ASERT, AST10-DLCMX)
• Microwave generator (Wattsine, WSPS-2450-200M)

Vacuum chamber

The vacuum chamber is a 6-inch 6-way standard ConFlat flange cube (Supplementary Fig. 16) used to provide an environment for the operation of the plasma thruster and plasma immersion ion implantation (PIII) into the palladium target. The vacuum chamber physically connects the plasma thruster and the electrochemical cell together. The vacuum system consists of a turbo pump, two dry scroll pumps, and a pressure gauge, and can reach a minimum pressure of 5×10⁻⁸ Torr. Additionally, the vacuum system is equipped with a residual gas analyzer to monitor the types of residual gas in the vacuum chamber. The vacuum chamber pressure is maintained between 1×10⁻⁵ Torr and 2×10⁻⁵ Torr during beam loading. The lowest pressure of the vacuum chamber during the electrochemical loading was approximately 4×10⁻⁵ Torr. The detailed components of the vacuum chamber, along with their make and model, are as follows:

• Dry scroll pump (Agilent, IDP-7)
• Turbo pump (Agilent, TwisTorr 305 FS)
• Pressure gauge (Edwards, WRG-S-DN40CF)
• Turbo controller (Edwards, Turbo Instrument Controller)
• Residual gas analyzer (SRS, RGA100)
• Plasma thruster holder (IV, SS, SWIFT-SEAL P1011088)
• Manual gate valve (KJL, SS, GV0400MVCF)
• Pneumatic bellows sealed angle valve (KJL, SS, SA0150PVCF)
• Vacuum chamber (KJL, SS, CU6-0600, 6-inch CF UHV Cube)

Electrochemical cell

The housing of the electrolysis chamber was custom-made from Macor (Machinable Glass Ceramic) and machined in-house using lathes and mills. A Viton O-ring holds the palladium target in place, physically separating the interface between the electrolyte and the vacuum. A brass rod, traveling the length of the electrochemical cell through a channel, connects the palladium target to the high voltage power supply needed to drive fusion reactions (Supplementary Fig. 17). The detailed components of the electrolysis chamber, along with their make and model, are as follows:

• Electrochemical cell holder (Taobao, SS, KF50 to 50 mm Compression port)
• Spring probe (QA Technology, 100-PLN1609L)
• Viton O-ring (McMaster-Carr, 1284N116)
• DC-DC converter (Walfront, Buck-Boost ZK-4KX)
• Lead-acid battery (Zeus, PC5-12F1-5)
• Remote control switch (eMylo, R121A)
• High voltage power supply (JIAMAN, H2105N-30-17)

Since the target is held at a high negative voltage (minus 30 kV), the electrochemical reaction is driven by a floating galvanostat. The galvanostat consists of a DC-DC converter and is powered by a 12 V lead-acid battery. The galvanostat is set to drive the electrochemical reactions at 200 mA between the cathode and anode for the electrochemical reaction (Supplementary Fig. 18).

In addition to connecting the galvanostat, the brass rod is also connected to a negative high-voltage power supply to drive the fusion reaction. The current value recorded by the high-voltage power supply is returned to the PID system as feedback data for automatic control of the PIII current.

Fusion reactions in the Thunderbird Reactor

The annealed and cleaned Pd target was installed into the electrochemical cell, which was then installed into the vacuum chamber. Vacuum was applied to the vacuum chamber until a pressure of less than 5×10⁻⁶ Torr was reached. Once a high vacuum was reached, deuterium gas was supplied to the plasma thruster at a flow rate of 0.5 sccm by a mass flow controller. The reactor control software was started after the mass flow controller read a steady flow rate of 0.5 sccm. The microwave generator was turned on with an input power of 10 W. The gate valve to the vacuum chamber was closed, and the vacuum increased from 5×10⁻⁶ Torr to 1×10⁻² Torr (we found that reduced vacuum conditions make the plasma ignition easier). The microwave generator power was increased to 200 W to ignite the plasma, which was verified visually by observing a light at the reactor pinhole opening (Supplementary Fig. 19). The gate valve was then opened to the vacuum chamber. The vacuum chamber pressure stabilized between 1×10⁻⁵ to 2×10⁻⁵ Torr. The PIII thruster current was controlled through a PID loop, which was set to 0.5 mA, with lower and upper thresholds of 0.3 mA and 0.6 mA, respectively.

The background rate of neutron count was collected for 5 minutes using the CAEN COMPASS software. The reactor was then started by applying minus 30 kV to the palladium target using a high voltage power supply. When the high negative voltage is applied to a target immersed in plasma, electrons are repelled, generating a plasma sheath (Supplementary Fig. 19). The plasma sheath, an electron-depleted region, appears darker than the plasma jet because it is difficult for ions to recombine with electrons. This sheath has a strong electrostatic field. As ions from the plasma jet enter this region, they are accelerated by the sheath voltage and injected into the target. A 6.67 mm pinhole is placed at the exit of the plasma thruster during the experiment. The purpose of the pinhole is to limit the PIII current to the 0.5 mA set for the reactor conditions.

For a beam loading experiment, the neutron production rate was collected for approximately two hours. In a typical experiment, a stable-state (when the rate of neutron production is no longer increasing as a function of time) was reached after approximately 30 min. We note steady-state as the time when all data points remain within ±5% of their mean value over a 30-minute period after the high voltage is applied.

For an electrochemically-enhanced experiment, the procedure was the same as above, except that the electrochemical cell was turned on after approximately 1 hour of reactor operation. The electrochemical cell was operated until the stable state was observed after the initial increase in the rate of neutron counts. The vacuum chamber pressure increased over time during operation of the electrochemical cell, and the final pressure was usually between 3.5×10⁻⁵ and 4×10⁻⁵ Torr.

Electrochemistry in the Thunderbird Reactor

All electrochemistry in the Thunderbird Reactor was conducted galvanostatically at a total current of 200 mA. A combination of a 12 V battery and a DC-DC converter was used to power the electrochemical cell. The electrochemical cell was manufactured from a cylindrical piece of Macor (machinable glass ceramic) with a diameter of 50 mm, a length of 265 mm, and a 1 cm diameter opening on the bottom facing the vacuum chamber. The palladium target, 300 μm in thickness with a geometric surface area of 1.5 cm² and exposed to the electrolyte, served as the cathode. The palladium target was sealed with an O-ring at the bottom of the electrochemical cell, with a geometric surface area of 0.785 cm² exposed to the vacuum chamber. The anode was an iridium wire, 320 mm in length and 0.5 mm in diameter. The iridium wire was rinsed with deionized water, dried with a kim wipe, and heated with a propane torch to eliminate adsorbed water. The cleaned iridium wire was placed into the electrochemical cell at a distance of 1 mm from the cathode. The electrochemical cell was filled with 17 g (13 mL) of 2 M K₂CO₃ in D₂O. The galvanostat was connected to the Pd target and the iridium wire using alligator clips to complete the electrochemical setup.

Electrochemical cycling in the Thunderbird Reactor

The effect of power cycling of the electrochemical cell was tested by turning the galvanostat on and off. An electrochemically-enhanced experiment was conducted as described above, with the electrochemical cell being activated after 60 min of beam loading, and the neutron production rate stabilizing at 172.6(4) n/s. The electrochemical cell was then turned off and on, in intervals of 30 minutes, while the beam loading of the Pd target with plasma continued.

Data availability

The data supporting the findings of this study are available within the paper and its Supplementary Information files. CAD files associated with the current submission are available at a repository DOI withheld until publication.

Code availability

All code for data cleaning and analysis associated with the current submission is available at a repository DOI withheld until publication. All code folders contain a detailed readme file to explain their use.

Acknowledgements

This project was supported by the Thistledown Foundation, Natural Sciences and Engineering Research Council of Canada (RGPIN-2018-06748), Canadian Foundation for Innovation (229288), and Canadian Institute for Advanced Research (BSE-BERL-162173). S.H. is supported by the Japan Society for the Promotion of Science Overseas Research Fellowships. We gratefully acknowledge Yet-Ming Chiang, Jeremy N. Munday, David K. Fork, Ross Koningstein, and Matthew D. Trevithick for their significant contributions to enabling the design of the Thunderbird Reactor, and to Ali Jalil, Aidan Madokoro, Adrien Noble and Daniel Pan for their assistance in constructing the reactor and in situ XRD cells.

(The preprint’s 42-item reference list, author affiliations and contribution statement are omitted here; the complete text is at the source.)