Tungsten oxide, Prussian blue, and some other metal compounds can go through a reversible electrolytic redox reaction that changes their color or transparency; commonly this involves intercalating lithium ions into them. But of course you can also electrolytically oxidize silver to black silver oxide or, if sulfur ions are available, to sulfide, and then reduce it again by reversing the current; this sort of thing is also done by artisans to add contrast to copper objects, typically using liver of sulfur rather than electrolysis.
This suggests that by running, say, silver or copper “electrochromic” strips in one direction and thin wire counter electrodes over them in the other direction, filling the space between with a thin electrolyte (maybe a hydrogel), you could get a very simple electrochromic display. It might not be fast or last through many switching cycles but it should still be interesting. Like other electrochromic displays it would be fairly bistable and thus potentially very energy efficient for passive reading.
If you apply higher voltages to speed up the reactions, unless you use a per-pixel diode or transistor, you might get some bleedover into other pixels in the same row and column, as well as the rest of the display. If you’re applying +2.1 V to a pixel, then any pixel not in the same row or column is in series with a pixel in the same row and a pixel in the same column with +0.7 V, -0.7 V, +0.7 V respectively. By a similar route, unless there are per-pixel diodes, different pixels will tend to drive currents through one another even when the driver is open-circuit, which will tend to equalize the charge and therefore the colors along each row and column.
With either per-pixel diodes or per-pixel transistors, the idea is that one of the two electrodes (let’s say the counter electrode, though the electrochromic electrode would work too) is divided into one section per pixel. In the diode case, there are two insulated wires for that row or column, one with a diode from it to the electrode, which can thus make the electrode an anode, and the other with a diode from the electrode to it, which can thus make the electrode a cathode. Ideally these would be germanium diodes, Schottky diodes, or both, to reduce the voltage error.
In the transistor case, the channel of a FET switchably connects the electrode to a power-supply line, which itself can be brought low or high, so you still have two insulated wires but you no longer have a voltage error. We’re using low enough voltages that the FET body diode probably doesn’t matter; if it does, you might be able to use a silicon carbide MOSFET (which has a larger body diode forward voltage because of carborundum’s 3.3 V bandgap, triple silicon’s; the MSC040SMA120B4 is rated for -4.0 V but the plot shows appreciable body-diode current at only -1.5 V, depending on Vgs) or I think you can get MOSFETs where the body terminal is brought out as a fourth pin, in which case you could tie that to a third power supply wire. (However, the 4-pin discrete MOSFETs I’ve been able to find use the fourth pin as a Kelvin-connection probe for sensing the voltage at the source on chip.)
The electrolytic reactions at the wire counter electrodes must also be taken into account; if they produce gas, for example, it will deplete the electrolyte, mechanically stress the device with gas bubbles, and may create an explosion risk. If the “wires” are, for example, transparent ITO strips, anything that forms on their surface will also be in the optical path; alternatively they could be the same metal as the electrochromic electrode, though they will probably have different overpotentials due to smaller surface area and thus higher current density.
You need the electrolyte to be on the same order of thickness as the pixel width in order to change the color of the whole pixel, though if the reaction passivates or “polarizes” the electrochromic electrode it might just be a question of how soon the color changes in each part of the pixel. That effect could be used to get, in effect, multiple pixels per intersection: whatever part of the electrochromic electrode is closest to the counter electrode would react first.
It may be useful to have reference electrodes that run along either rows or columns in order to control the voltage on the electrochromic electrode more precisely.
Such a device could presumably be used as short-term nonvolatile memory as well, using the thickness of the passivation layer thus formed to record a bit, measured by the ratio of resistive impedance to capacitive impedance by probing it at two frequencies.
Some materials have different extinction coefficients (opacities) for different wavelengths, so the color of their films depends on their thickness, quite aside from iridescence. For oxide layers that are not very opaque at any wavelength, the iridescence effect will tend to be stronger than the inherent color of the oxide formed, though it will be weaker in contact with water than with air, since the index of water is 1.33, close to common glasses. However, zinc oxide is 2.4, hematite is close to 2.9, tenorite is 2.9-3.1, titania is 2.6, and the strength of the reflection at the interface is roughly proportional to the square of the difference of the indices, so such materials would still have great potential for iridescence.
In general these devices will act faster at higher temperatures.
The Pourbaix diagram for copper shows that above about pH 7 and above about +0.3 volts the equilibrium favors black cupric tenorite, CuO; as pH increases to about 12.5 the critical voltage decreases to about -0.2 volts. But there’s a small region, for example from about -0.1 V to about +0.2 V at pH 8, where instead red cuprite, Cu2O, is favored. (Different sources disagree on exactly how big this window is.) At more negative voltages, the equilibrium favors the reduction back to copper metal.
In this case the electrolyte would need to be slightly alkaline, and maybe you could get three colors: copper yellow, red, and black. Possibly turning a pixel red might take weeks.
There also exists an unstable olive-green copper peroxide, but I don’t think you can make it in this way; you need pre-existing peroxide groups.
If the copper forms dissolved copper salts, they will of course be green, and when it redeposits as metallic copper it will often be yellow rather than shiny. Oxides of copper are very insoluble, though, so this presumes some other materials in the electrolyte.
Copper oxide itself is an electrochromic material and when it contains some cuprite it is reported to be somewhat reddish-gray even when only 60-500 nm thick.
Iron oxides can have many different colors, especially with water hydroxylating them: in pottery commonly red, green, grey, or brown; there are sixteen known oxides, including black Fe3O4 magnetite, black FeO wüstite, red Fe2O3 hematite, orange/brown FeOOH goethite which can be yellow to black depending on things including limonite at the yellow end, and green. This is another possible multicolored pixel, although you probably can’t get all of those colors; the Pourbaix diagram for iron in water at 25° says that starting about pH 8.1, you get iron up to about -0.5 V, (green?) fougèrite up to about -0.3 V, black magnetite up to about 0 V, red hematite up to about 1.2 V, and then aqueous ferrate solution, “pale violet... one of the strongest water-stable oxidizing species known” (!). However, I suspect that most of these reactions are very slow.
Nickel is pretty passive most of the time, but nickel oxide is used in pottery to produce blue, grey, yellow, and black, and its usual NiO form is green, while the trivalent oxide-hydroxide is black. I’m not sure if you can form the divalent green compound in water; the Pourbaix diagrams I’m finding are contradictory.