CONTROL OF RESIDUAL ANALYSES.
Of the three factors to be determined in the residual air, the oxygen (which is most important from the standpoint of the relative weight to be placed upon the analysis) unfortunately can not be directly determined without great difficulty. Furthermore, any errors in the analysis may be very greatly multiplied by the known errors involved in the determination of the true volume of the air in the chamber as a result of the difficulties in obtaining the average temperature of the air. Believing that the method of analysis as outlined above should be controlled as far as possible by other independent methods, we were able to compare the carbon dioxide as determined by the soda-lime method with that obtained by the extremely accurate method used by Sondén and Pettersson. An apparatus for the determination of carbon dioxide and oxygen on the Pettersson principle has been devised by Sondén and constructed for us by Grave, of Stockholm.
In the control experiments, the air leaving the mercury valve D (fig. 30, page 66) was caused to pass through a T-tube, one arm of which connected directly with the sampling pipette of the Sondén gas-analysis apparatus, the other arm connecting with the U-tubes for residual analyses. By lowering and raising the mercury reservoir on the gas-analysis apparatus, a sample of air could be drawn into the apparatus for analysis. The results of the analysis were expressed on the basis of moist air in volume per cents rather than by weight, as is done with the soda-lime method. Hence in comparison it was necessary to convert the weights to volume, and during this process the errors due to not correcting for temperature and barometer are made manifest. However, the important point to be noted is that whatever fluctuations in composition of the residual air were noted by the soda-lime method, similar fluctuations of a corresponding size were recorded by the volumetric analysis with the Sondén apparatus. Under these conditions, therefore, we believe that the gravimetric method outlined above is sufficiently satisfactory, so far as the carbon-dioxide content is concerned, for ordinary work where there are no wide variations in the composition of the air from period to period.