The Akhilleus-heel of the spark-gap closing switch is its high jitter, due to the stochastic nature of streamer formation and cosmic-ray bombardment in the working gas. By stimulating the emission of photoelectrons from a surface with a low work function, you can lower the breakdown voltage of a spark gap significantly, enough to get it to trigger reliably with low jitter; if you can drop the breakdown voltage by more than about 25%, you should be able to reliably trigger a discharge with little risk of prefires (and, in addition to dropping the breakdown voltage, this photoemission also significantly reduces the variability). But, historically, the only reliable way to get the short-wavelength light necessary to stimulate such photoelectron emission was either an incandescent object or another electrical discharge, leading to a chicken-or-egg problem.
LEDs in the 365–400-nm range are now widely available, including US$10 high-power 3-watt jobbies, and for US$4.56 Digi-Key will sell you a 278nm 1-milliwatt Everlight ELUC3535NUB-P7085Q05075020-S21Q UVC LED designed for UV sterilization. LEDs normally have no jitter and can have rise times measured in nanoseconds, though the larger ones have enough junction capacitance to slow this down, and the time lag of photoemission is typically also around a nanosecond.
To reduce the threshold frequency of light necessary to provoke photoemission from the spark-gap cathode, we can coat it with a low-work-function surface; I think barium oxide is the traditional choice for this in vacuum tubes. In a spark gap you would need to ensure that the air within was dry enough not to allow the barium oxide to become wet. Zinc might be a friendlier metal, but it still must be protected from oxidation — or nitridation! This suggests you might need a controlled atmosphere in the spark gap.
I think hν is the energy of a photon, where h is Planck’s constant 6.63 × 10⁻³⁴ J/Hz, and ν is the frequency; and the threshold frequency is that at which this energy is the work function of the surface. This gives 3.1 eV at 400 nm, 3.4 eV at 365 nm, and 4.5 eV at 278 nm.
Barium metal’s work function varies from 2.52–2.70 electron volts on different crystal faces, putting it within the grasp of even the 400-nm “debatably ultraviolet” LEDs, though its enthusiastic reactivity is problematic; also thus reachable are sodium at 2.36 and lithium at 2.9, cerium at 2.9, and maybe yttrium at 3.1. The more sedate cerium and yttrium are perhaps more promising, though they are pyrophoric and quickly oxidize in air, the oxides are passivating.
The 365-nm LEDs might additionally be able to spall photoelectrons off manganese at 4.1 eV and neodymium at 3.2, and the 278-nm ones could bring within reach zinc at 3.63–4.9, lanthanum at 3.5, molybdenum at 4.36–4.95, and even tin at 4.42 and lead at 4.25. Unfortunately all the metals whose oxides are less stable in air than the metals themselves (gold and some of the platinum group) have work functions that are still out of reach.
Barium oxide formed in a certain way on a silver substrate has a work function around 3.2 eV, and the barium peroxide (which BaO tends to turn into at room temperature, given the chance) is up around 3.6. On tungsten, barium oxide mixed with oxides of strontium and calcium lowers the work function below 2 eV, and baria alone is calculated to be 2.7 eV. Magnesia, much more chemically stable than baria, has apparently also been used to good effect; although by itself its work function is 4.22–5.07 eV, a thin film of it on a metal surface apparently reduces the work function? I don’t know.
So, this suggests a setup with a hermetically sealed gas-filled spark gap where the anode has one or more holes in it through which an ultraviolet LED can shine onto the cathode; the cathode has a partial coating of one of the above systems, such as a thin film of barium oxide on top of a coating of titanium, or a coating of cerium, a coating of lead, a coating of zinc, a coating of lead-tin solder, or a coating of tin. When the spark gap is held a little below its breakdown voltage, a pulse of current through the LED can initiate abundant photoemission into the interelectrode gap, lowering the breakdown voltage enough that the spark gap triggers without any voltage change, and with a jitter measured in nanoseconds.
The advantage of having many holes is that a larger area of anode is exposed to the photoelectron-enriched region of the gas, potentially permitting higher current. The advantage of not having many holes is that lower arc inductance can be achieved by having many parallel arcs around the edges of a single round hole, and it’s easier to fabricate.
All of these same techniques can also be applied to a pseudospark switch, and of course any other low-work-function material can also be used. Pseudospark switches normally have jitter down in the tens of ns.
Although UV irradiation drops the breakdown voltage, I’m not sure it drops it below the lowest safe non-UV-irradiated breakdown voltage. If that is the case, this approach will always have a significant chance of prefires. (I don’t know why free electrons in the gap don’t drop the lowest breakdown voltage, but apparently free ions in the gap from a previous firing do.)
A hybrid approach, however, should work extremely well: use an UV LED to illuminate a conventional electrically-triggered spark gap, using a third triggering electrode (whether insulated with quartz or not). This should give jitter that’s as low as could be hoped for with the LED, while eliminating the risk of prefire. You still need a low-work-function electrode surface to keep the work function low enough to overcome with a mere LED. (Hofstra sells a UV-illuminated spark gap using this principle, but I think it uses conventional a mercury-discharge-lamp UV source rather than LED illumination; it appears to be hand-sized.)
There are reports that in the twilight zone below the Paschen minimum, where pseudospark switches operate, electron injection is adequate to reliably trigger a discharge, and “UV flash” is an existing triggering approach. Perhaps electron injection via UV-LED-induced photoelectrons would be sufficient. Normally pseudospark switches operate at neon-sign-style vacuums (10–50 Pa) in order to get past the Paschen minimum, but you could instead simply make them very small (micron-sized gaps), reducing the distance factor rather than the pressure factor.
To avoid arcing to the low-work-function “seed” surface — for example, if it’s delicate, has annoyingly high resistivity, or would contaminate the dielectric gas undesirably — it can be placed closer to the anode than the main cathode is, connected to the rest of the cathode with a heavy high-value resistor. As long as only the photocurrent is flowing in the gap, the voltage across the high-value resistor is effectively zero, but once the spark initiates to the seed, the seed quickly reaches the potential of the anode, so the spark will rapidly propagate to the rest of the cathode through the gas, since now the entire gap’s breakdown voltage is across the much shorter distance between the two parts of the cathode, and across the resistor.