July 28, 1898] 



NATURE 



305 



more than uniformity in the conditions and in the method 

 of experiment. For these several reasons, in translating old 

 temperature observations into the normal hydrogen scale, it 

 would in general be a waste of labour to aim at the degree 

 of accuracy possible in the best thermometric work. 



Whilst the exact determination of other scales in terms of the 

 hydrogen scale is from a historical standpoint less important 

 than might appear at first sight, it is still rendered essential by 

 the fact that for many practical purposes the hydrogen ther- 

 mometer is inconvenient, and is unlikely to supersede other 

 forms. 



The most exact scale comparisons are doubtless those made at 

 Sevres, under the auspices of the international committee of 

 weights and measures. These are described, with the exception 

 of the most recent, in Dr. Guillaume's " Thermometrie de 

 Precision," a work which all really interested in exact ther- 

 mometry should study for themselves. Air being a composite 

 medium, and so presumably less suitable for the basis of an 

 exact scale than the elementary gases, has apparently not been 

 dealt with at Sevres ; but the work there has included the com- 

 parison of the hydrogen, nitrogen and carbonic acid scales, espe- 

 cially of the first two. The investigation covered, in the first 

 instance, the range -25° C. to +100° C. , and was executed 

 with great care by Dr. Chappuis. In point of time it preceded 

 and, in fact, led up to the adoption of the hydrogen scale. The 

 comparison of the gas scales was not direct, but through the in- 

 termediary of mercury thermometers. From the data on p. 258 

 of Guillaume's "Thermometrie," one learns that within the 

 range 0° to 100'' C. the difference between the hydrogen and 

 nitrogen scales does not exceed 0° "Oil C, but at - 25° C. it 

 amounts to about o°"ol6. The differences between the hydrogen 

 and carbonic acid scales are five or six times as large as these. 

 The hydrogen temperature is algebraically less than the nitrogen 

 or carbonic acid temperatures between 0° and 70° C, but alge- 

 braically greater at temperatures below 0° C. In Guillaume's 

 opinion it is probable that ordinary (constant volume) air ther- 

 mometers give a scale near that of nitrogen, but lying somewhat 

 on the side of the carbonic acid scale, i.e. more remote from 

 hydrogen. The probable error in Dr. Chappuis' comparisons 

 is given as ±o°"OOi between 0° and 50° C. , and twice or thrice 

 as great at either +75" C. or -25° C. 



The differences between the several gas scales presumably 

 increase as the temperature falls, but probably never become 

 large. At all events, in 1896 Holborn and Wien {Wied. Ann., 

 vol. lix., 1896, p. 213), using constant volume thermometers 

 (with, however, an initial pressure of only one atmosphere at 

 o°C.), found the hydrogen thermometer to read only about o°'6 

 C. higher than the air thermometer at - 190° C, a temperature 

 close to the freezing point of air. Ten years earlier Olszewski 

 found a difference of about 1° between the hydrogen and 

 nitrogen scales at - 150° C. ; but his thermometry was probably 

 less exact. With only Olszewski's results before him, Guillaume 

 infers that the hydrogen scale is almost certain to agree closely 

 with the absolute scale, even at - 220" C. ; and Holborn 

 and Wien's observations lead them to a somewhat similar 

 conclusion. Recent comparisons by Olszewski of hydrogen 

 and helium thermometers (Nature, vol. liv. pp. 378 and 544) 

 are strongly confirmatory. 



For every-dayuse, unfortunately, gas thermometers are some- 

 what cumbrous. The international committee accordingly as- 

 signed an important place in their programme to the determin- 

 ation of the relations between the hydrogen scale and that of 

 the glass-mercury thermometers which they have selected as 

 working standards. The thermal expansion of glass, though 

 small, is not negligible compared to that of mercury, and varies 

 in different kinds of glass. The international committee ac- 

 cordingly selected one special kind of glass, French verre diir, 

 as standard. The selection of the glass does not alone suffice 

 to fix the scale. No glass has yet been discovered whose 

 behaviour is decided wholly by the existing conditions. 

 When a thermometer after exposure to a temperature of 

 50° C. is placed in ice, it reads lower than it would 

 have done prior to the exposure, and this depression of 

 zero, as it is called, increases to a certain extent with the 

 duration of the previous heating. It is thus necessary for 

 high accuracy to decide on a uniform plan of dealing with 

 this source of uncertainty. The plan adopted by the inter- 

 national committee is to refer every reading of a thermometer to 

 a zero determined immediately after the reading. Under cer- 

 tain circumstances enough is known of the behaviour of verre 



NO. 1500, VOL. 58] 



dur to permit of the substitution for the ac(ual zero observation 

 of results extracted from a table of zero depressions. After a 

 reading is taken with a verre dur thermometer, a variety of 

 corrections have to be applied. These are necessitated by 

 inequalities in the bore or errors in graduation, by the influence 

 of the external pressure exerted by the atmosphere and the in- 

 ternal pressure exerted by the mercury. Verification at Sevres 

 consists in evaluating and tabulating all the necessary corrections. 

 After these corrections are applied, the result represents the 

 temperature on the natural verre dur — mercury scale. ^ This 

 scale has been compared with that of the hydrogen thermometer 

 at Sevres from about -38°C. to -h2CX5°C. Below - 10° C. 

 and above 100° C. the comparison is probably less exact than 

 between these limits. 



In considering the probable accuracy of temperature measure- 

 ments made with verre dur thermometers, we have to take into 

 account the consistency of readings taken with the same ther- 

 mometer, the closeness of readings taken under the same 

 conditions with different thermometers, and, from certain points 

 of view, the degree of accuracy with which readings can be 

 reduced to the hydrogen scale. 



The consistency of readings taken with a single verre dur 

 thermometer depends in the first instance to some extent on the 

 success with which the correction tables have been constructed 

 at Sevres ; it varies to a large extent with the skill of the 

 observer, the conditions of the experiment, and the temperature 

 to be measured. The ordinary verre dur standard thermometer 

 is divided to o°*l C. and read by estimation, with the aid of a 

 lens magnifying from ten to twenty times, to o°"OOi C. This 

 involves subdivision of a space into hundredths by eye, a feat 

 which the skilled observers at Sevres accomplish with marvellous 

 accuracy, but which is far beyond the powers of the ordinary 

 experimenter. In some instances use can be made of a 

 micrometer, but this can hardly be employed unless temperature 

 is practically stationary ; and, when this is the case, troubles 

 are apt to arise from capillary action in the mercury. The more 

 remote the temperature to be measured from that of the sue-, 

 rounding air, the greater, as a rule, is the probable error of an 

 observation. Thus, speaking generally, observations between 

 0° and 40" C. are those capable of the highest accuracy ; and 

 here it would appear that the mean results obtained on different 

 occasions by skilled observers for a fixed temperature with a 

 verre dur thermometer may be expected to agree to within 

 about + '001° C. 



At temperatures below ioo°C. the corrected readings of different 

 verre dur thermometers on the same occasion show apparently 

 about as good agreement as is to be expected from the readings^ 

 of a single verre dur thermometer' exposed on different occa- 

 sions to the same fixed temperature. At temperatures, however, 

 approaching 200° C. Dr. Chappuis found that the corrected read- 

 ings of different verre dur thermometers might differ by as much 

 as o°05 C. 



The accuracy with which the relation of the verre dur to the 

 hydrogen scale is known is hard to say. Until the Sevres com- 

 parisons have been repeated at other places, under equally 

 favourable conditions, there will always remain a certain amount 

 of doubt as to the existence of possible local or temporary in- 

 fluences. In the meantime, it is not altogether reassurmg that a 

 recent partial comparison of the hydrogen and verre dur scales 

 at 10°, 20°, 30° and 40° C, by Dr. Chappuis, gives results differ- 

 ing from those of the original comparison at Sevres by from 

 o°ooi to o°-oo7 C. 



From the above data two considerations naturally arise. At 

 temperatures between - 20° C. and 100° C. the natural verre 

 dur scale is probably that most easily and exactly realised in 

 practice ; and it is, perhaps, fully as correct to regard the present 

 normal hydrogen scale as one deducible in a prescribed arbitrary 

 way from the verre dur scale as to accept it as having any real 

 physical existence. On the other hand, there is no such thing as 

 a verre dur scale, unless we agree to neglect differences of tem- 

 perature which are of the same order as differences actually found 

 between different verre dur thermometers. At the present moment, 

 for instance, we must apparently treat o°-05 C. as a negligible 

 quantity in temperature measurements at 200° C. if we are to 

 extend the verre dur scale to that point. It has to be borne in 

 mind that identity in the chemical constitution of thermometer 

 glass may not necessarily imply identity in temperature scale. 



1 For any natural glass-liquid scale, thermometet d<?gree divisions include 

 equal volumes of the bore. 



