370 PROCEEDINGS OF THE AMERICAN ACADEMY. 



not only possible, but extremely probable. A difference of one tenth of 

 one per cent in the volume of nitrogen makes a difference of 0.3 in the 

 value of the atomic weight. The errors of collection and transference 

 of the gas are more likely to result in reading too small rather than too 

 large volumes, giving too high values for the atomic weight. 



From these considerations, it is evident that Aloy's results are at least 

 somewhat doubtful. Aloy gives notice of his intention to apply this 

 method to the determination of other atomic weights, but it is to be 

 hoped that before doing so he will clear up some of the doubtful points 

 in connection with the process. As carried out in this investigation, the 

 method certainly is not a valuable addition to the methods of atomic 

 weight determination. 



From the earlier results Clarke computed the value 239.6, while the 

 German Committee recommend 239.5. Both figures are practically 

 identical with Zimmermann's figures. 



The investigation herein described was undertaken with the hope that 

 by increasing the experimental basis of our knowledge of the subject, we 

 might be able to reduce to somewhat narrower limits our present uncer- 

 tainty in regard to the real value of this constant. 



Preliminary Work upon the Preparation, Properties, and 

 Methods of Analysis of Some Uranium Compounds. 



In view of the well known advantages of the halogen compounds for 

 accurate analysis, when these compounds can be prepared and weighed 

 iu a state of purity, — it seemed desirable to use a halogen compound as 

 the basis of a determination of the atomic weight of uranium. 



Of the four chlorides of uranium known to exist, none can be pre- 

 pared in a state of purity that is beyond question. Green uranous 

 chloride, UC1 4 , which results from passing dry chlorine over a mixture 

 of uranium oxide and carbon at red heat, is easily converted to the 

 pentachloride, UC1 5 , by further action of chlorine at high temperatures. 

 There can be no positive evidence that the green chloride would not 

 contain some of the pentachloride, and if the attempt is made to prepare 

 the pentachloride from the green chloride, it is equally difficult to be sure 

 that the conversion is complete. The trichloride, UC1 3 , is made by reduc- 

 ing the tetrachloride with hydrogen, and here again it is difficult to be 

 sure that the tetrachloride is completely reduced. Uranyl chloride, 

 UOoCL, cannot be prepared in the dry state. 



It is extremely probable, then, that any of the chlorides will contain 

 larger or smaller quantities of a higher or lower chloride. It may be 



