With the Baldwin experimental locomotive 60,000 (1926), designed for high efficiency, evaporation declined from 10 to 6.5 pounds water per pound of dry coal, as firing rate increased from 30 to 150 lb/ft2 grate/hr. and superheat increased from 180oF to 257oF. (Pennsylvania RR, Fig. 19).
If you burn carbon in air, the hydrogen will be contaminated with nitrogen from the air. This can be avoided by burning pure oxygen into carbon monoxide, but then you must provide the oxygen somehow.
The process can be operated on a mostly continuous basis; occasionally clinker must be removed. (Teel 81). Water gas has impurities, such as hydrogen sulfide and ash (84).
Water gas in turn can undergo this shift reaction, discovered by Felice Fontana in 1780:
CO (28 grams) + H2O (18 grams) -› CO2 (44 grams) + H2 (2 grams)
Since exposure to CO (carbon monoxide) is dangerous, naturally there was interest in conducting the steam-carbon reaction in such a manner as to minimize its formation, i.e., to obtain the mixture of carbon dioxide and hydrogen:
2H2O + C -› CO2 + 2H2.
Gillard found that this could be accomplished by use of an excess of steam. The carbon dioxide can be removed on a batch basis (see below), but unfortunately it proved "very difficult to carry this out in practice on a large scale. . . ." (Sander).
BAMAG worked at a low temperature (at which the reaction equilibrium is favorable), but with catalysts (typically nickel) to speed up the reaction. This results in what is reportedly the cheapest method of producing hydrogen (1 shilling/9 pence per 1000 cubic feet), but unfortunately the product contained 4% nitrogen, a serious disadvantage for aeronautical use. (Greenwood 162). 2002McGHEST suggests a reaction at 350oC over an iron oxide catalyst.
Griesheim-Elektron instead disturbed the water gas equilibrium by "absorbing" the carbon dioxide with lime or other alkali. Cost of production (1912) was 2s/2s.5p-2s/9p per 1000 cubic feet for a moderate size plant. While the process can be carried out at a lower temperature than the steam-iron process below, reducing maintenance costs, "the handling of the large amounts of lime presents some difficulty." (Greenwood 167ff).
Of course, we can eschew the shift reaction, and remove the carbon monoxide with an "absorbing agent" or by liquefaction. EB11/Carbon notes that it is "rapidly absorbed by an ammoniacal or acid (hydrochloric acid) solution of cuprous chloride," but the resulting hydrogen is only 80% pure. (Sander) Later, Frank and Caro thought of employing heated calcium carbide. This conveniently "absorbed" not only carbon monoxide, but also carbon dioxide and nitrogen, and in the process produces graphite and calcium cyanamide. (Sander; Elis 597).
Liquefaction (Linde-Frank-Caro method) at -200oC works well, but small-scale plant costs are high (Ellis 460) and concerns have been expressed about the dangers of working with compressed carbon monoxide (595). In 1912, a plant producing 3500 cubic feet hydrogen/hour cost about 13,000 pounds, and had a cost of hydrogen production of 3 to 4 shillings per 1000 cubic feet. (Greenwood 174).
A little more explanation of liquefaction may come in handy. A gas can only be liquefied if cooled below its critical temperature; at that temperature, it must be compressed to the critical pressure; at lower temperatures, lesser pressures are needed for condensation.
(Teel 114ff).
It can be seen that cooling water gas to -200oC (usually by surrounding it with liquid nitrogen boiling under reduced pressure) permits separation at normal pressure. Or one may use a more moderate cooling and greater-than-atmospheric pressure. The liquefied carbon monoxide is used to pre-chill the incoming water gas, and then is burnt as a fuel. (Teel 119).
Steam-iron (dry) process. Lavoisier was the first to show (1783) that hydrogen could be produced in an acid-free reaction, by reacting iron with steam. And Argand recognized that the yield would be higher than with the acid process. (Clow 159).
There are really two different steam-iron processes, the single step non-regenerative one for field use, in which iron is consumed, and the two-step cyclic one, in which iron oxide is reduced with water gas to iron, and then the iron is reoxidized to iron oxide with steam, for large plants. The former has the simplified equation
Fe ( 56 grams) + H2O (18) -› H2 (2) + FeO (72)
Steam-iron generators were used by the world's first air force, the 1794 Compagnie d'Aerostiers, as this process was safer, cheaper, and most important, because the sulfuric acid was needed for gunpowder manufacture. (Boyne 378; Langins 536).
In autumn 1793 Coutelle placed iron scrap into a three foot long, one foot diameter cast iron reactor pipe, which in turn was placed inside a furnace. Water was introduced into the pipe, and turned to steam (Langins 537). There's gas flow rate data for six experiments: 0.32-1.114 (0.544 average) m3/hr. Water flow rates were in the 22-58 g/min range. (539). In one experiment, in the course of four days and three nights, 23.83 cubic meters hydrogen was produced. (538). That corresponds to about 40% yield.
The scale of these experiments was almost 150 times that of in the 1780s, and a scale-up problem was encountered: carbon dioxide. The wrought iron musket barrel used in the laboratory experiments was essentially free of carbon, but Coutelle's cast iron pipes were up to 4.5% carbon. The carbon dioxide would reduce the buoyancy of the gas; the density of the produced gas ranged from one-half to one-sixth that of air; pure hydrogen would be about one-fourteenth. The problem was addressed by bubbling the gas through lime water, but even then the gas was on the heavy side (possible explanations include failure to change the limewater frequently enough; use of too high a gas flow rate; or failure to first remove dissolved air. (549).
In March 1794, the technology was taken to a new level; each of seven reactor pipes, eight feet long, one foot in diameter, and one inch thick, were stuffed with 540 pounds iron. (This was essentially the system taken into the "field," see Hoffmann 24). Note that each pipe, so loaded, weighed a ton. At this new scale, a problem that had been minor before became more significant: if the reactor pipes were heated too vigorously; they cracked or melted (cast iron melts at 1050 oC). (543). Sometimes the heating was uneven, with some pipes softening and others not heated enough (546). It didn't help that the pyrometers in use were poorly calibrated, and the pipes were poor in quality; locally produced pipes were even permeable to water! (547). There were also problems with the cement used to seal the reactor pipe to the exterior tubing; it cracked and thus gas was lost. (546, 550).
These French steam-iron generators are best viewed as "relocatable" rather than truly portable field equipment. After the 1793 experiments, it was envisioned that a balloon could be filled five or six days after the arrival of the generator equipment at the launch site. (Langins 542). In actual practice, the furnace, with a twenty-foot tall chimney, took twelve days to build (16,000 bricks were needed, and local bricks weren't necessarily refractory enough). (546). In contrast, the American Civil War generators were up and running within a matter of minutes.
With these main generators, it reportedly took 50 hours to fill a 30 foot diameter (4500 cubic feet) balloon (Delacombe 31) and such production was considered slower than the acid-iron method. (Boyne 378).
However, the French did also have a smaller scale furnace that had a single pipe and could be assembled in an hour. It used as much fuel as the large furnace, but could produce 800 cubic feet in 48 hours using 180 pounds of iron. It was used to "top off" an inflated balloon that had leaked while transported from the main facility to the launch site. (Langins 552).