328 Messrs. Haldane and Penibrey on an Improved Method 



added, and the bottle allowed to stand, with occasional 

 shaking, for at least two hours. The baryta-water was then 

 removed to a clean and dry cylinder, corked up, and titrated 

 after the carbonate had subsided. We used as a rule either 

 30 or 40 cub. cent, of the baryta-water in the titration, which 

 was performed twice with the ordinary oxalic-acid solution 

 (1 cub. cent. ='001 gram, of C0 2 ), rosolic acid or phenol- 

 phthaleine being used as an indicator. The end reaction was 

 always sharply defined, and the two results never differed by 

 more than O'l cub. cent. 



The above Table shows that with air containing a small 

 proportion of carbonic acid the Pettenkofer method gave 

 results from 8 to 27 per cent, higher than those given by the 

 gravimetric method. As the proportion of carbonic acid 

 rises (Experiment No. 9) , the difference tends to disappear. 

 Judging from these experiments the result for outside air 

 would usually be about 15 or 20 per cent, too high by the 

 bottle method. 



It will be seen that the pairs of determinations made with 

 the gravimetric method agree with the utmost closeness, both 

 for moisture and carbonic acid. 



We have not succeeded in tracing satisfactorily the reasons 

 for the inaccuracy of our results with the bottle method. 

 Although these results were all too high, other observers 

 have, as before mentioned, obtained results which were often 

 almost certainly too low. 



We at first thought that the error might be due to contact 

 of the baryta-water with air during the manipulations in- 

 volved. In the first four or five analyses we had simply 

 followed Pettenkofer's plan of pouring the baryta-w r ater out 

 of the large bottle through a funnel into the narrow vessel 

 used for allowing the carbonate to settle. To avoid the 

 exposure to air implied in this operation, we in the later 

 experiments removed the baryta-water with a pipette pro- 

 vided with a piece of rubber tubing for applying suction. 

 As the pipette was washed first with some of the baryta, it 

 was free from carbonic acid. Our results, however, were 

 little, if at all, better. 



Another possible source of error is the state of the glass 

 w^alls of the bottle. Pfliiger * found in making the titrations 

 for Kjeldahl nitrogen determinations that it was impossible 

 to titrate accurately in flasks which had previously been 

 washed with acid. The glass seemed to retain some of the 

 acid in spite of washing, and give it up again slowly during 

 the analysis. We made several experiments with a bottle 

 * Pfliiger's Archiv, xxxvi. p. 105. 



