﻿926 Mr. A. H. Davis on Natural 



the heat loss per unit area from corresponding points is 

 given by 



hcck0{0 9 ac/kvyt*, (13) 



and this will also be true if h represents mean heat loss for the 

 whole model. So for bodies of this shape and orientation 

 we have 



h = k0(6ffac/kvy' 3 f(L*0gac/kv), . . . (14) 



which may be written 



hL/k0 = ¥(U0ffac/kv). ..'.-. (15) 



This equation is the simplified form of equation (1) it was 

 desired to establish, and it has been put to the test of 

 experiments in a later part of this paper. It is desirable to 

 notice here one point in connexion with it. 



For a series of fluids for which cvjk is constant, the 

 equation may readily be shown to agree with that obtained 

 by Boussinesq for inviscid fluids ; i. e. Boussinesq's grouping 

 of variables for invisicid fluids is satisfactory for viscid 

 fluids for which cvjk is constant. This equivalent grouping 

 is given by omitting (cvjk) from formula (1). 



Part II. — Experimental. 

 Formula?. 



For long horizontal wires of diameter ' d ' it may readily 

 be shown that formulae (1) and (2) may be rewritten as- 

 follows in terms of the heat loss H per unit length of wire 

 per degree temperature excess : 



Klk=Wgd*a6/0)Acv/k),) (m 



R/k=F(d 5 0ffac/kv), J 



When cvjk is a constant, the equations are identical in 

 form and, consequently, evidence for diatomic gases already 

 shown elsewhere to be in agreement with one of* these 

 expressions is necessarily in agreement witli the other. 



If cvjk is not constant, the second equation is a special case 

 of the first. The experiments now to be described on the 

 cooling of wires in liquids will indicate the form of the cvjk 

 term in (16) and also whether the simpler expression is 

 satisfactory. 



Apparatus. 



The method of experiment consisted in stretching a wire 

 horizontally at a convenient depth in a vessel full of the 

 liquid under examination, and measuring the electric energy 



