REPORT ON THE COMPOSITION OF OCEAN-WATER. 163 
I was quite dissatisfied with these irregular results, and need not say that I did not 
by any means feel sure that Bunsen had gone even further wrong than myself. But not 
being able to discover any flaw in my work serious enough to account, for instance, for the 
fact that while Bunsen (for the temperature 4°'5) makes 1000 \ = 22 , 08, I find it to be 
26 '7 (i.e., 4 c.c. per litre more), I thought there was no use in simply repeating my 
determinations, and, instead of doing so, started a series with pure nitrogen gas, prepared 
by sucking a slow current of dry air free from carbonic acid over a long column of red- 
hot copper wire-gauze into a Pisani # gas-holder, charged with a very dilute solution of 
alkaline pyrogallate. As my Jacobsen flask held 800 c.c., I needed for each experiment 
several litres of gas. We did not always succeed in keeping such large supplies of nitrogen 
gas absolutely free of adventitious air. The nitrogen, accordingly, was viewed as a mix- 
ture of pure nitrogen with a little oxygen; and the formulae on page 162 employed for 
calculating the result. After the saturation of the water with the nitrogen had been 
effected, a sample of the latter was put aside for determining m l should the analysis of the 
absorbed gas give more than a practical nil for n- L . As the values n x were always very 
small, the values of m 1 could be calculated from those of n 1} because we evidently have — 
g' 1 = m 1 /3 1 P , 
q 2 = m 2 /3 2 P , 
where the symbols q l q 2 designate the quantities of oxygen and nitrogen absorbed by the 
water operated upon from the impure nitrogen with which it was saturated. As an 
obvious sequence from the two equations, we have — 
A = 
A + A 1 
and, since approximately, /3 y = 2/3 2 , we may say — 
or since is small, 
n x — 2toj- 
1 
+ m i 
n x = 2 m 1 . 
In this manner I generally calculated n x and m 1 from each other, and for each adopted 
the mean of the value calculated and the value found. The greatest value for m l which 
ever occurred was 0‘0079. When n x was found to be very small, «q was merely calculated 
from it, and not determined at all. 
* Two bottles, provided with cork-holes below, and through these united by means of an india rubber tube. One 
of the bottles is provided with a glass stop-cock, inserted by an india-rubber stopper in the neck. 
* 1 
