272 



NATURE 



[July 20, 1882 



cal considerations : more or less weight must be given to ! 

 different results on other than purely mathematical 

 grounds ; hence identical final results would not always 

 be arrived at by different calculators starting from the 

 same experimental data. Inasmuch as "the atomic | 

 weight of each element involves the probable error of all 

 the other elements to which it is directly or indirectly 

 referred," it may happen that the probable error attach- 

 ing to an atomic weight determination is large, although 

 the experimental data are extremely accurate. Thus, 

 Crookes, by very accurate experiments, found the atomic 

 weight of thallium to be 203 '642 ; but this number sup- 

 poses that N0 3 = 6r889; the value to be now assigned 

 to the atomic weight of thallium depends on the accu- 

 racy with which the atomic weights of oxygen and nitrogen 

 have been determined. The work of Crookes simply 

 fixes, with great accuracy, the ratio between the equiva- 

 lents of Tl and N0 3 . 



The most probable value for the atomic weight of 

 oxygen is found to be 1 5 '9633 (H = 1), with a probable 

 error of ±'0035; any error which there maybe in this 

 determination is involved in the determinations of the 

 atomic weights of most of those elements which come 

 after oxygen. When the atomic weight is large, the error 

 thus introduced may be considerable : thus if 0=15-9633 

 Ur = 238-482, but if O = l6 - oo Ur = 239-03 ; difference 

 = 0-548. 



Some of the weighings involved in the calculations have 

 been reduced to absolute standards, others are only un- 

 corrected weighings in air ; hence an error is sometimes 

 introduced which cannot be eliminated. 



The discovery of Dumas that silver prepared by the 

 method of Stas, occludes weighable quantities of oxygen, 

 has been already referred to ; in four experiments Dumas 

 found that 1 kilogram of silver occluded S2, 226, 140, and 

 249 milligrams of oxygen respectively; the largest of 

 these numbers is taken by Prof. Clarke as "Dumas' 

 correction." The effect of apphing this correction is 

 generally very slightly to lower the value of the atomic 

 weight ; the following table exhibits this effect in a few 

 instances : — 



Uncorrected. 



Silver 107-923 



Chlorine ... 35 '45' 



Bromine ... 79-751 



Iodine 126S4S 



Potassium ... 39*109 



Sodium ... 23"o5i 



Corrected. Difference. 



107-896 ... - -027 



35'4/S ... +-027 



79-978 ... +-027 



I26S75 ... +-027 



39-oS3 ... --026 



23-024 ... --027 



In the appendix is given a table containing the mean 

 atomic weights (with probable errors atliched) for all the 

 elements, calculated from the most trustworthy data. It 

 is shown that twenty-five out of the sixty-six elements 

 considered have atomic weights the values of which differ 

 by less than one-tenth of a unit from whole numbers, 

 (H = 1) but many of those numbers which differ by more 

 than this fraction involve any error which there may be 

 in the determination of the value of the atomic weight of 

 oxygen, multiplied many times. If the possible error in 

 the value for oxygen be transferred to that for hydrogen, 

 i.e. of O = 16, then it is shown that forty-four out of the 

 sixty-six elements have atomic weights differing by less 

 than one-tenth of a unit from whole numbers. Of these 

 forty-four elements, twenty-six show plus, and thirteen 

 minus variations from whole numbers. Those which 

 exhibit minus variations are discussed in detail ; the 

 values for the atomic weights of seven of these have not 

 been determined with any great accuracy ; silver alone 

 has a value which carries "very much weight against the 

 hypothesis of Prout." Cf those elements, twenty-six in 

 number, the atomic weights of which exhibit plus varia- 

 tions (less than o'i) from whole numbers, three — viz. 

 Nb, Yt, and Ur — have values which have been very in- 

 accurately determined ; seven involve " Dumas' correc- 

 tion," the application of which will bring the values 



nearer whole numbers (Al, As, Ba, Cd, Li, P, and Na). 

 Special sources of possible error are indicated in the dis- 

 cussion of the atomic weights of Al, Ca, F. Five, of the 

 twenty-six elements, have atomic weights the values of 

 which involve errors due to the possible occlusion of 

 hydrogen by the metals when reduced from their com- 

 pounds (Co, Fe, Ir, Ni, and W). In many other cases 

 the variations from whole numbers are extremely small, 

 i.e. much less than one-tenth of a unit. " In short in the 

 majority of instances the errors may be diminished by 

 corrections which are in all probability needed, and 

 which we can easily point out." 



Twenty-six elements have atomic weights the values of 

 which vary more than one-tenth of a unit from whole 

 numbers ; of these twenty-six, three — viz. CI, Rb, and Sr 

 have values nearly half multiples of that of hydrogen ; 

 the atomic weights of nine — viz. Sr, Au, In, La, Rh, Ru, 

 Si, Te, and Zr — have been very imperfectly determined ; 

 the atomic weights of Sb, Ce, Be, Yt, Pt, and Hg are 

 discussed, and it is shown that the atomic weights of 

 these elements may come within Prout's hypothesis ; no 

 criticism is offered on the atomic weights of Cr (52-129 ± 

 •025), Cu (63-3181-011), Mo (95747 ± -051), and V 



(5 1 '373 ± '°-4)- 



The value to be assigned to iodine (126848 ± 022) 

 depends on that for silver ; at present iodine stands as an 

 important exception to Prout's rule. Potassium presents 

 a serious objection; but if "Dumas' correction" is ap- 

 plied K becomes 39-083 [O = 16]. Clarke concludes by 

 saying that although he began his examination of atomic 

 weights strongly prejudiced against Prout's hypothesis, 

 the facts have obliged him to give it " a very respectful 

 consideration." " AU chemists must at least admit that 

 the strife over it is not yet ended, and that its opponents 

 cannot thus far claim a perfect victory." 



The recalculation of atomic weights shows clearly to 

 the chemist what experimental work ought now to be 

 undertaken ; revisions of the atomic weights of tellurium, 

 silicon, boron, mercury, chromium, manganese, uranium, 

 and gold are urgently called for. The " periodic law " 

 requires that the atomic weight of tellurium should be 

 smaller than that of iodine ; although the mean number 

 recently obtained by Wills is greater than 127, yet this 

 number cannot be accepted as final. Several results 

 brought out by Clarke have an important bearing on the 

 "periodic law." In most of the tables of elements 

 arranged in accordance with the law, didymium is placed 

 before cerium and lanthanum ; Clarke however shows 

 that Di = 144-573 (± 031) ; Ce = 14042+ (± '°'7) ; a «d 

 La = 138-526 (± 03). Brauner, in his paper recently- 

 published in Chem. Sac. Journal, finds Di = 14618 (mean 

 of three results), and La = 13958 (mean of two results). 

 We may therefore conclude that La<Ce<Di. These 

 elements then come in series 8; lanthanum in group III. 

 giving the characteristic oxide La 2 3 , cerium in group IV. 

 giving the oxide Ce 2 4 , and didymium in group V. with 

 the oxide DijOg, which oxide has been lately prepared 

 and examined by Brauner {Chem. Soc. Jul. Trans., 1882, 

 p. 68). 



Prof. Clarke's work may be taken as a type of what is 

 now so much wanted in chemistry : a careful collection 

 and digestion of masses of facts. We seem to be for- 

 getting that chemistry is a science, not a collection of 

 facts. Every week adds fresh material to the heap ; the 

 scien:e is in danger of being crushed beneath the load of 

 details. M. M. Pattison Muir 



FIRE RISKS FROM ELECTRIC LIGHTING 



AVERY strong and influential committee was recently 

 formed bv the Society of Electricians to draw up a 

 series of rules and regulations not only for the guidance 

 and instruction of those who have electric lighting appa- 

 ratus installed on their premises, but for the reduction to 



