As for transistors at GHz frequencies...
http://www.cree.com/news-media/news/article/cree-introduces-highest-power-and-frequency-plastic-packaged-gan-transistors-for-low-cost-radar-and-datalinks

Quote from http://www.cree.com/news-media/news/article/cree-introduces-highest-power-and-frequency-plastic-packaged-gan-transistors-for-low-cost-radar-and-datalinks

Smaller RF Transistors are Ideal Replacements for GaAs IMFETs and Commercial Tubes in Amplifiers with Power Levels up to 100W Continuous Wave and Operating Frequencies up to 10 GHz

That's direct operation up to 10 GHz... double that with a step recovery diode (or any 'snappy' diode) and VIC.

There are liquid lasing mediums. They generally produce light in the near-IR to near-UV range. In fact, water has been used as a lasing medium, as well... except the slower oscillations of the water molecule result in longer wavelengths than visible light.

https://phys.org/news/2016-11-water-wave-laser.html

But I get your point... we're essentially trying to use the WFC as a lasing medium container, and pushing it to the point of ionizing the lasing medium.

Rather than trying to force the water molecule apart by brute force, let's look at this from a different perspective... that of Rydberg state coupling.

A Rydberg state is sort of the opposite of the Bohr orbit... rather than describing the electron orbit starting at 1s and going up, it describes the difference between the electron energy and that electron's ionization energy.

There are theoretically 32 excited states of water, so that rather narrows the search for that state which will induce autoionization upon excitation. Rydberg state coupling in some of those excited states leads to autoionization.

Quote from http://www.phys.ttu.edu/~gglab/rydberg_state.html

It starts to get really interesting when you consider an energy region where there are states associated with more than one channel present. These different channels might, for instance, be associated with different spin-orbit states of the ion core in an atom or different rotational and vibrational states of a molecular ion core. Additional interactions which have been neglected in the definition of the channels give rise to couplings between the states of the different channels, leading to complicated energy shifts and intensity variations in the Rydberg series. If, in one or more of the channels in the energy range of interest, the excited electron is unbound (open channels), this channel interaction leads to a mixing of the bound states (in closed channels) with the unbound states and the process of autoionization.

Rather than trying to hit the water with light... how about electron impact dissociation? Electrons can carry far more energy than light can, and we've had the technology for 148 years to create an electron beam.

We'd easily reach the energy levels necessary (a typical color TV CRT will generate ~20,000 eV), we can easily control the energy level of the electrons, and we could even steer the electron beam if need be.

It should be said that ionization efficiency declines with increasing electron energy due to the de Broglie wavelength of the electron becoming smaller than the water's bond length... so we'd start at low energies and work our way up.

The average O-H bond length (per diffraction studies) is 1.01 angstrom, which corresponds to 1.484 eV. This is the minimum energy needed if you hit the water molecule at the exact right angle. And of course, it's dependent upon water temperature and pH. It's been reported there's a dissociation peak at ~16 eV.

You'll note the electron kinetic energy is approximately 1000 times less than the required photon energy... the electron's mass (and hence kinetic energy) imparts far more energy than a similar-energy massless photon.

The trick is to get the electron energy just right such that we minimize radiation collisions (which would just heat the water) and maximize ionization collisions. One way of doing this is by running the electron beam through water vapor rather than bulk water... remember, the smaller the water droplet size, the lower the ionization barrier.

So running an electron beam through OH^- or H₃O⁺ rich water vapor consisting of very tiny droplets should give a significantly reduced ionization barrier.

What happens to the electron after all its kinetic energy is absorbed by the water droplets (it should give off Bremsstrahlung radiation the whole way... making the water vapor glow... something we want to try to minimize, as that's a radiation collision rather than the desired ionization collision)? It comes to 'rest' in the water at the particle range, causing the water to build up a negative charge... "charged water".

Because the water vapor droplets have a net negative charge, the droplets will repel each other, preventing their agglomerating into larger droplets.

As for aligning the water molecules in a static magnetic field, that strengthens the inter-molecular bonds, which causes the O-H bond to strengthen as well, so it'll be slightly higher than 1.484 eV, but you gain more from hitting the water molecule at just the right angle than you lose from strengthening the O-H bond. That is dependent, of course, upon magnetic field strength... the stronger the magnetic field, the shorter (stronger) the water's bond length, but the higher the likelihood that the water molecule is properly aligned. Since a permanent magnet has a divergent magnetic field, you'd have to mount a permanent magnet and Helmholtz coil outside each side of the tube carrying the water vapor droplets. The coils would 'steer' the magnetic flux so it's not divergent. But if you could align the water molecules and hit them with an electron beam at just the right angle, you're looking at ~9 times less energy required to dissociate the water.

As regards the tank I described in an earlier post... there's really no need to put the HV electrodes in the water itself. If the container is made of a dielectric material, the HV electrodes can be placed outside the container, and the water ionized by field strength of the high voltage alone... thus no current would flow. The low voltage electrodes would still be in the water, to prevent reassociation of the OH^- and H₃O⁺ ions, but that voltage would be below the ionization overpotential, so very little current would flow when using pure water, so long as the neutral-pH low-conductivity central region of the tank's water remained sufficiently wide.

Think of the overpotential (energy barrier) to ionizing water as a ramp. A DC voltage or an electrostatic field strength applied across the water will raise the lower end of that ramp. A high enough applied DC voltage or electrostatic field strength will tip the lower end of the ramp above the other end of the ramp (the energy barrier), and you get electrolysis (and current flow in the case of the DC voltage). We want to avoid doing that, to prevent current flow.

Water being right on the edge anyway (it constantly swaps protons, undergoing auto-ionization then reassociation), we want to raise the voltage or field strength just enough such that the inherent random fluctuations which cause that auto-ionization are at very nearly the same 'ramp level' as the energy barrier, then prevent reassociation via the low voltage separating the cations and anions.

The same applies for a magnetic field... a strong enough magnetic field will cause water to ionize... but the field strength necessary is so high as to be impractical.

As for plate material, if you're going to immerse the HV electrodes, the HV (+) plate can be stainless steel. In an oxygen-rich environment, stainless steel develops a chromium oxide layer (although you'd have to clean the plate prior to using it to remove any machining chemicals and cutting tool iron deposits, as they inhibit passivation layer formation... so a long soak in nitric or citric acid and a good wash with pure water is called for), which is a dielectric.

You'll note that due to hydrogen adsorption occurring if the plate is used with AC current, and due to that hydrogen adsorption making the chromium oxide layer much more conductive, stainless steel should only be used for a pure-DC or pulsed-DC system, and only for the (+) electrode. You definitely want to keep any free hydrogen atoms away from that chromium oxide passivation layer.

You'll also note that in an oxygen-deprived environment, stainless steel degrades pretty quickly, because it can't form and maintain that passivation layer.

For the low-voltage DC electrodes (used to segregate the OH^- and H₃O⁺ ions), stainless steel isn't going to be your best choice because of that passivation layer inhibiting ion segregation.

Thus, you might think about using Monel as your low-voltage DC electrodes and the HV (-) electrode. It's very close in the Galvanic Series to stainless steel, so you won't get a lot of galvanic corrosion, and it's highly resistant to corrosion and acids, which you'll need for the HV (-) plate anyway, since that will have the H₃O⁺ rich water, which will be acidic.

Or, you can use Monel for all the plates to preclude any possibility of galvanic corrosion, and figure out some way of coating the HV plates with a dielectric such as polyester.

In the big picture the process always appears to follow Maxwell's demon theory in that a discrete series of actions which only occur on the atomic level in themselves simply do not incur the same losses as bulk actions on large volumes. Like a little Maxwell demon ninja who simply slices H2O apart into H2 and O2 and does not allow them to recombine incurring all the losses we have considered to be normal.

It seems fairly simple and elegant in it's approach and discrete individualized actions are not bulk actions any more than apples are oranges.

As for the high voltage required, pure water has a dielectric strength of 65-70 MV/m, dependent upon electrode material work function and work function differential between the electrodes (in the case of different materials being used for each electrode), electrode surface, and electrode surface area. Thus for every millimeter width of central-region neutral-pH low-conductivity water we have, we want ~65kV. We can tailor the width of the central-region neutral-pH low-conductivity water by adjusting the voltage... lower voltage gives a wider region, higher voltage gives a narrower region.

So to maximize the pulsed HV effect of suddenly ionizing the water molecules (then the low voltage would shuttle them away from each other to prevent recombination), we'd want to have electrodes which are "pointy" on their surface... in other words, a flat smooth electrode is going to give the worst performance, whereas an electrode surface populated with many small pointed spikes will give the best performance, even if that electrode surface is then covered in a dielectric.

Think in terms of surface charge density here... the reason lightning rods are pointed is because that presents a higher surface charge density to any potential lightning that may develop, 'steering' the lightning toward the lightning rod. The same concept applies here... the high voltage we put on the HV electrodes can be lower (versus a flat, smooth electrode) if we surface the electrode with many tiny sharp points on its surface, maximizing the charge density at the tip of each point.

We can further enhance the HV effect by utilizing the 'water hammer' effect inherent in a rapidly opened switch. We could use something like a triggered spark gap, as in a Marx Generator, if the circuitry such as Meyer used is unworkable.

When I was playing around with a circuit emulator, I cobbled together an analog to a Marx Generator without realizing it. With a 12 volt input, it produced sparks (about 4 spark events per second) in the low MV range.

As for electrode placement, we want to separate the anions and cations via high voltage, then shuttle them away from each other via low voltage to prevent recombination.

One way of doing this is by separating the anions and cations into different planes by putting the low voltage plates at each end of the tank, and the high voltage plates in the middle of the tank at the top and bottom.

So the process would go roughly something like this:
We spike high voltage into the tank. This separates the anions and cations in a vertical plane in the center of the tank. So for instance, the OH^- will be higher than the H₃O⁺.

Then the low voltage shifts those cations and anions horizontally toward the low voltage plates. Because of the vertical separation of the anions and cations, they have less probability of encountering an opposing ion and recombining, which would only result in heating the water and wasting the energy used to separate them.

This has the problem of water level dropping and uncovering the upper high voltage plate... unless the plate was placed such that the tank water level maintenance system ensured it'd never be uncovered, this could represent a hazardous condition... high voltage arcing in a hydrogen and oxygen rich atmosphere.

Another way is to have the same polarity low and high voltage plates at one end of the tank, with the opposing polarity low and high voltage plates at the other end of the tank. This has some problems if you put the high voltage plates outboard (more toward the end of the tank) of the low voltage plates, in that the high voltage must go through the already ionized OH^- and H₃O⁺ rich water. This puts that region closer to electrolyzing the water.

Remember, we don't want to electrolyze the water, we want to separate the anions and cations, then recombine them in the engine. We can safely store the anions and cations, we can't as safely store or handle hydrogen.

There are also problems with placing the high voltage electrodes inboard of the low voltage electrodes... namely that the material of the high voltage plates will impede anion and cation movement toward the low voltage plates... unless perforated or mesh electrodes are used.

And the final method is to use the same plates for both the high and low voltages. This requires more sophisticated electronics which cuts off the low voltage, applies the high voltage, then cuts off the high voltage and applies the low voltage. One has to contend with the capacitance of the plates holding that high voltage for a brief time... so your low voltage circuitry has to be able to handle high voltage, too.