689 



The index of refraction n of the surface ia^er will increase on 

 approaching the reflecting metal surface. When for n a mean value 

 is taken, viz. that between ^iaó- and the value for the greatly condensed 

 air, immediately adjoining the metal and when with Quincke') it is 

 supposed, that the density of this is equal to that of the mercury, 

 n = 4:.04:8 is found for this b}^ the aid of the relation: {n — l):d = 

 constant'). Hence the mean value of n is (4.048 + 1.003) : 2 = 2.52. 

 As A' — A, the phase difference for the absorbed layer of air, amounts 

 to 0.0087 (see § 6) and (p = 79°46' (see § 6), we find with the 

 values of / and H mentioned in § 8? 



L ^ i.Q fi(x. 



This value is in agreement with that for the transition layer 

 liquid-vapour, for which Bakker^) gives J — 2 fi,a. 



When in the same way the thickness of a layer of oil (72 = 1.5, 

 c/=0.9) is calculated, which according to ^ 5 can modify the angle 

 of principal incidence by 1°, the compensator- read ing by 0.50 and 

 more, we find, introducing the value 0.50, L = 3 ft/L/. This is in 

 accordance with Rayleigh's and Fischer's^) determinations. Rayi-eigh 

 found, namely, for the thickness of the thinnest layei' of oil, that stops 

 the movements of the camphor particles on water, 2/ifi. F'ischer 

 found for liquid layers, which spread over mercury, thicknesses 

 smaller than 5/i/Lt. 



As the adsorbed layer of air of a thickness of 1.6 mi changes 

 the compensator-reading by 0.25 and the mean error in the reading 

 amounts to 0.02, a layer of a thickness of 0.13 n^i can still be 

 demonstrated in this way by this optical method. Such a layer is 

 of the thickness of a molecule. It is not possible to prove the 

 existence of such thin layers by the aid of the capillary phenomena. 



It is not possible to remove the once adsorbed layer of air by 

 means of a very far exhausted vacuum, as the mercury airpump of 

 Gaede can bring about. After eight hours' pumping no displacement 

 of the compensator reading could be demonstrated *). 



1) Quincke, Pogg. Ann., 108, 326, 1859. 



*) L LoKENZ— H. A. LoRENTz's formula cannot be applied here, as n^ would 



M,- + 2 n' + 2 nH2 d 



become negative. Let — di be = —z — - d, then — ; must be > 1, ifWi' 



n^ — 1 n — 1 71 — 1 d^ 



d, n' + 2 



IS to be positive. In the case considered here , = 10*, — - — - =5.10', so that the 



d ?r — 1 



condition is not fulfilled. 



3) G. Bakker, Z. f. Phys. Chem., 91, 571, 1916. 



4) Rayleigh, Phil. Mag., (5), 30, 396, 1890; Fischer, Wied. Ann., 68, 436, 1899. 

 ») This result is in agreement witli other experiments, which show with how 



45 

 Proceedings Royal Acad. Amsterdam. Vol. XXI. 



