The French military ballooning company was abolished in 1799, and the steam-iron process didn't return to public notice until 1861. Then, John Wise proposed that the Union adopt a horse-drawn generator wagon with two eighteen-inch cylindrical retorts made from boiler plate, a boiler, and a firebox, the latter generating heat for both steam generation and for heating the solid reactants (a combination of charcoal and iron turnings, so this was really a blend of the steam-carbon and steam-iron methods) in the retorts. (I visualize this as a bit like a locomotive, but with retorts replacing the piston cylinders.) A tender was to carry water, iron turnings, and firewood, again much like a locomotive. He asserted that a 20,000 cubic foot balloon could be inflated in four hours.
Contemporary calculations showed that for a single inflation, the wagon would need to carry 7500 pounds of iron turnings, and twenty-two cubic feet of water (1364 pounds by my calculation). The machinists declared that the generator would weigh more than five tons, and of course the tender would be additional. The estimated cost of the generator was $7,000. The government declined to fund the project. (Haydon 77).
The British experimented with the steam-iron method in 1879. Captain Templer reportedly generated hydrogen at a rate of 1,000 cubic feet/hour (BLE 109). The apparatus weighed 3.5 tons, was carried in three general service wagons, and could generate enough gas for two balloons in twenty-four hours. "But the apparatus did not prove satisfactory." (Baden-Powell 742). Possibly, the British expectations were too high; Templer said that he wanted a production rate of 5,000 cubic feet/hour. By 1885 the British had switched to shipment of compressed hydrogen to the field. (Moedebeck 226).
Insofar as field use in the 1632 universe is concerned, we clearly aren't going to want to go with an installation like that used in the French Revolution. The question is whether a viable system could use a firebox and boiler similar in size and design to that of a nineteenth-century steam locomotive, and then route the steam through reactor tubes to oxidize the iron. (Coutelle in fact considered use of steam engine cylinders as reactor tubes-the steam engine existed in 1794, even though the steam locomotive didn't.) Such a system could at least be transported by rail.
From the equation quoted earlier, it takes nine pounds of water to make one pound of hydrogen, and 5.23 pounds hydrogen occupies 1000 cubic feet (20oC). A pound of coal should evaporate five to ten pounds water (although more coal would be needed to heat the iron and to superheat the steam to the most effective reaction temperature). But it seems to me that the steam-iron process on the locomotive scale should work (although not necessarily better than the acid-iron process).
One thousand cubic feet hydrogen provides about seventy-two pounds of lift. And to produce it, you need one hundred ten pounds of iron. So carrying iron on board an airship for hydrogen production at destination is a losing proposition.
The purity achievable with the early-twentieth century embodiment of the steam-iron process is 98.5-99% (Taylor 172) . The forward reactions are:
2H2O + 2Fe -› 2FeO + 2H2
3H2O + 2Fe -› Fe2O3 + 3H2
4H2O (72 grams) + 3Fe (168) -› Fe3O4 (232) + 4H2 (8)
The reaction products of the simple process do have possible utility; FeO as a black pigment, Fe2O3 as a red pigment and as jeweler's rouge, Fe3O4 as a black pigment and a catalyst in the water gas shift and other reactions. And, of course, all can be smelted to regenerate iron. Which, of course, is one stage of the regenerative process.
Getting the regenerative steam-iron process working properly isn't trivial. The iron-producing reduction with water gas is endothermic and the hydrogen-producing oxidation is exothermic.
It may be possible in a large plant to use waste heat from retorts that are in the steaming step stage to warm retorts that are in the water gas stage. However, that heat isn't enough, by itself. (Taylor 55). Both the carbon monoxide and the hydrogen of water gas are reducing agents, but there are fuel economies in working at lower temperatures, which favor carbon monoxide activity. The catch is that this results in higher levels of carbon and carbon monoxide in the next step. (Greenwood 178)
Temperature has several different effects. Higher temperatures shift the equilibrium point in favor of the reverse reaction, but the reaction is forced forward by continuously removing hydrogen and supplying fresh steam (Greenwood 275). The permeability of the iron to steam and hydrogen is reduced by fritting above 900oC. However, the reaction is very slow below 650oC. (Teel 88).
In the Lane multi-retort system, dry steam is used at a pressure of 60-80 psi without any superheat; the exothermicity of the reaction raises the reaction temperature adequately. In a single retort system, the same pressure is used with partial superheat. (Taylor 51).
Ideally, the "contact mass" of iron is porous (to maximize reactive surface area) yet robust (so it doesn't crumble into dust and create a back-pressure). Also, it is resistant to local overheating, which results in sintering. The choice of iron (spathic ore is best) makes a difference; "before 1917, American producers . . . imported their contact material, mainly from England." (29). Spongy iron-manganese ores will prove to work better than ordinary iron ore. (Ellis 495, 502). It may be possible to catalyze the reaction with copper, lead, vanadium or aluminum. (Ellis 502).
As iron oxide is formed, it shields the remaining iron from the steam. (Teel 87). Hence, for large-scale economical operation, the iron (which was oxidized to iron oxide) is regenerated. (Ellis 485). That means that you may start with iron ore (Fe2O3) instead of iron. About six tons iron ore are needed to produce 3500 cubic feet/hour. (Teel 88). The iron oxide is reduced (probably with water gas), and then you introduce the steam to react with the iron and produce the hydrogen. Then you repeat the cycle. Typically, consumption of water gas is 2.5 cubic feet per cubic feet hydrogen in a multi-retort plant and 3.5 in a single retort one. (Teed 97).
The water gas, in turn, is produced by the steam-coal process discussed earlier. Soft coke consumption is at a rate of about one ton for every 6500-7000 cubic feet hydrogen. (98).
However, the water gas must be purified or the impurities will result in formation of adverse deposits on the contact material or gaseous contaminants (hydrogen sulfide, carbon dioxide, etc.) in the hydrogen. (Ellis 487ff; Roth 25). There is also sulfur in the iron ore (Teel 93), and sulfur compounds are especially problematic. (Greenwood 181). Measures taken to cope with these problems increase the volume of water gas required and also reduce the production rate. (Ellis 487). Even then, the iron eventually loses its activity. (Ellis 499). "An unsuccessful attempt at commercial production . . . was made by Giffard in 1878. The iron rapidly became inactive due to sintering of the material and to chemical reaction with impurities in the reducing gases used." (Taylor 27).
Even in modern embodiments, the initial product contains a large fraction (61%) of steam; that can be condensed out. There will also be carbon monoxide (from the water gas) and nitrogen (presumably from dissolved air in the water used to make the steam). (Brewer 232). These are purified out.
The most common factory implementation of the regenerative steam-iron process is the Lane process; it's relatively economical of fuel but there's more deposition of carbon and (thanks to side-reaction with steam) higher carbon monoxide content. A plant producing 3500 cubic feet/hour might have 36 vertical retorts, each 9 inches diameter and 10 feet high. (Greenwood 178). Despite the recommendations of 2002McGHEST, the typical retort temperature was 650oC, prolonging the useful life of the retorts. They last 12-18 months, and the ore is good for 6. Water gas is consumed at rate of 2-3 volumes per volume hydrogen, and the cost of hydrogen is 3/- to 4/- per 1000 cubic feet, excluding overhead. (182).