594 PROCEEDINGS OF THE AMERICAN ACADEMY. 



rather than as a definite statement of fact, this risk of partial mistake 

 in some of the details can do no serious harm. 



In the first place, there are many well-known reasons for believing 

 that liquid water is partly polymerized, and that its polymeric form is 

 more bulky than its simple form. The admirable experimental work of 

 Ramsay and Shields on surface tension * and the interesting logic of 

 van Laarf may be cited as among the work affording a strong presump- 

 tion of this partial " association." The determination of even the average 

 extent of the polymerization is a more difficult matter, as may easily 

 be seen by studying the papers of van der Waals j and Ramsay § upon 

 the subject. A conservative estimate is perhaps that given by Ramsay 

 in his last mentioned paper — namely, the estimate that water at 0° has 

 an average molecular weight of 1.7 X 18, and at 140° an average 

 molecular weight of 1.3 X 18. Thus the liquid under ordinary condi- 

 tions seems to consist of a mixture of (H. 2 0)„ and H 2 0, the latter pre- 

 ponderating at high temperatures. || Presumably ice consists wholly of 

 (H,0) n . 



The determination of the magnitude of n in the above polymer is even 

 more difficult than the determination of the exact average extent of 

 polymerization ; but fortunately for the present discussion neither of 

 these points is crucial as far as it is concerned. Usually n is taken as 

 2 ; but there seem to me several reasons for believing it to be 3, an 

 assumption which would correspond to other familiar polymers and with 

 the well-known trigonal symmetry of ice.1I In this case, of course, less 

 of the polymeric form would be needed to account for the anomalous 

 properties of water, than if its polymerized formula were (II 2 0) 2 . 



In any event, because of the well-marked univalence of hydrogen, 

 every one will concede that the polymer is probably formed by the union 

 of oxygen to oxygen. The two most probable structural forms will then be 



II m H 



IL n II , H 00 H 



H 00 H and " O n 



II H 



Either assumption will satisfy the requirements of the present argument. 



* Ramsay and Shields, Zeitschr. phys. Cliem., 12, 433 (1893). 



t Van Laar, ibid., 31, 1 (1899). 



% Van der Waals, Zeitschr. phys. Chem., 13, 714 (1894). 



§ Ramsay, Zeitschr. phys. Chem., 15, 106 (1894). 



|| This conclusion does not exactly agree with the highly hypothetical as- 

 sumptions of Sutherland. Phil. Mag., (5) 50, 460 (1900). 



If The crystal form of ice is cited by Sutherland as a reason for assuming 

 (H,0) 3 . 



