In such a room, however, if the doors and windows were absolutely air tight, and there was no fireplace, diffusion through the walls would change the entire air once an hour, so that the percentage would not rise above 0.04; while in any ordinary room imperfect workmanship and an open chimney would change it four times in the hour, reducing the percentage to 0.01, a quantity which the most inveterate enemy of water gas could not claim would do more than produce a bad headache, an ailment quite as likely to have been caused by the same factor that brought about the blowing out of the gas.

Moreover, we are now talking about the use of carbureted water gas as an enricher of coal gas, and not as an illuminant to be consumed per se. and we may calculate that it would be probably used to enrich a 16-candle coal gas up to 17.5 candle power. To do this 25 per cent. of 22 candle power carbureted water gas would have to be mixed with it, and taking the percentage of carbon monoxide in London gas at 5 per cent. - a very fair average figure - and 18 per cent. as the amount present in the Van Steenbergh gas, we have 8.25 per cent. of carbon monoxide in the gas as sent out - a percentage hardly exceeding that which is found in the rich cannel gas supplied to such towns as Glasgow, where I am not aware of an unusual number of deaths occurring from carbon monoxide poisoning.

The carbureted water gas has a smell every bit as strong as coal gas, and a leak would be detected with equal facility by the nose; and I think you will agree with me that the cry raised against the use of carbureted water gas, for this reason, is one of the same character that hampered the introduction of coal gas itself at the commencement of this century.

We must now turn to the chemical actions which are taking place in the generator of the water gas plant, and these are more complex in the case of the Van Steenbergh plant than in those of the Lowe type, and, for that reason, yield a gas of more satisfactory composition.

Taking gas as made by the Lowe or Springer process, and contrasting it with the Van Steenbergh gas, we are at once struck by several marked differences.

In the first place the hydrogen is far higher and the marsh gas or methane lower in the Van Steenbergh than in the Lowe process, this being due to the sharper cracking that takes place in the short column of cherry red coke, as compared with the lower temperature employed for a longer space of time in the Lowe superheater. Next we notice a difference of 10 per cent. in the carbon monoxide, which is greatly reduced in the Steenbergh generator by the carbon monoxide and marsh gas reacting on each other as they pass over the red hot surface of coke with formation of acetylene, which adds to the illuminants, this action also reducing the quantity of marsh gas present.

In the illuminants, if we add the higher members of the methane series present to the olefines, we see they are about equal in each gas, while the low percentage of nitrogen in the Lowe gas is due to more careful working, and could easily be attained with the Van Steenbergh plant by allowing the first portion of water gas to wash out the producer gas before the hopper on top is closed.

The cracking of the naphtha by the red hot coke is undoubtedly a great advantage, for, as I have pointed out, the cracking of rushing petroleum is an exothermic reaction, so that the coke at the top of the generator gets hotter and hotter, and it is no unusual thing to see the coke at the beginning of the make cherry red at the bottom and dull red at the top, while at the end of the make it is almost black at the bottom and cherry red at the top, in this way attaining the same advantage in working that the Springer and Loomis do by their down blast, that is, having the fuel at its hottest where the gas finally leaves it, so as to reduce the quantity of carbon dioxide, and so lessen the expense of purification.

It will be well now to turn for a few moments to the gas obtained by cracking the light petroleum oils by themselves. The Russian and American petroleum differ so widely in composition that it was necessary to see in what way the gases obtained from them differed; and to do this, equal quantities of American naphtha and a Russian naphtha were cracked, by passing through an iron tube filled with coke, and in each case heated to a cherry red heat, the gases being measured, and then analyzed, with the following results:

Showing that, if the Russian oil is a little lower in illuminants, it quite makes up by extra volume, but it seemed to me to deposit a much larger proportion of carbon.

Taking 21/2 gallons of American naphtha, it would give roughly 180 cubic feet of gas of the above composition, while the remaining gas would be the ordinary water gas. Taking the analysis of this as given, and calculating from it what would be the composition of a mixture of it with the naphtha gas, we obtain:

Showing how great the effect is of the diluents in the water gas in preventing the overcracking of the hydrocarbons, as shown by the increase in the percentage of them present in the finished gas; while the enormous reduction in the amount of carbon monoxide present is due to the interaction between it and the paraffin hydrocarbons in the presence of red-hot carbon, a point which makes the Van Steenbergh apparatus enormously superior to any of the superheater forms of plant.

After all said and done, however, the reactions taking place, although they have an intense fascination for the chemist, are not the factors which the gas manager deems the most important, the cost of any given process being the test by which it must stand or fall; and it will be well now to consider, as far as it is possible, the expense of enriching coal gas by the various methods I have brought before you.

In order to be well above the prescribed limit of illuminating power at all parts of an extended service, the gas at the works must be sent out at an illuminating power of 17.5 candles and we may, I think, fairly take it that 16 candle coal gas, as made by the big London companies, costs, as nearly as can be, 1s. per 1,000 cubic feet in the holder, and the question we have now to solve is the cost of enriching it from 16 to 17.5 candle power. When this is done by cannel, the cost is 2.6 pence per candle power, so that the extra 11/2 would cost 4d. per 1,000.

Carbureting by the vapors of gasoline by the Maxim-Clarke process costs 13/4d. per 1,000, so that the extra candle power would mean an expenditure of 2.62d. Unfortunately I have no figures upon which to calculate the cost of producing such a gas by the Dinsmore process, but with the three important water gas enrichers we can deal.

Using Russian fuel oil, which can be obtained in bulk in London at 3d. per gallon, the proprietors of the Springer plant guarantee 51/2 candle power per 1,000 cubic feet of gas per gallon used, so that, to produce a 22 candle gas, 4 gallons would be used. The cost per 1,000 cubic feet may be roughly tabulated, as the coke used amounts to about 40 lb.

Twenty five per cent. of 12-candle gas when mixed with 75 per cent. of the 16-candle gas gives the required 17.5 candle gas, which would therefore cost 1s. 11/2d., or the enrichment would have cost 11/2d.

By the Lowe process, an increase of 5.3-candle power is guaranteed for the consumption of a gallon of the same oil, so that the cost would be a shade higher, all other factors remaining the same, while with the Van Steenbergh process both grade of oil and consumption of fuel vary from either of these processes. In order to obtain a thousand cubic feet of 22-candle gas, two and a half gallons of the lighter grade oil would be consumed, and I am informed that there is now no difficulty in obtaining oil of the right grade in London in bulk at 4d. per gallon, which would make the cost: