Investigations on Platinum Thermometry at Kexo Observatory .
3 " Investigations on Platinum Thermometry at Kew Observatory .
" By C. Chree , Sc. D. , LL. D. , F.R.S. , Superintendent .
Communicated by the Kew Observatory Committee .
Received December 5 , \#151 ; Read December 14 , 1899 .
Contents .
Sect. page 1 .
Preliminary ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... . .
3 2 , 3 .
Thermometers ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... . .
4 4 , 5 .
Resistance box and coils ... ... ... ... ... ... ... ... ... ... ... . .
5 6 .
Nature of the observations ... ... ... ... ... ... ... ... ... ... ... . .
7 7 , 8 .
Sources of change or error ... ... ... ... ... ... ... ... ... ... . .
8 9 , 10 .
Changes in the " leads " or in the " proportional arms " ... .
10 11 , 12,13 .
Coil changes ... ... ... ... ... ... ... ... ... ... ... ... ... ... 13 14 .
Faulty action of the contact piece ... ... ... ... ... ... ... ... . .
16 15 .
Shift of the bridge centre ... ... ... ... ... ... ... ... ... ... ... .
16 18 .
Thermo-electric currents ... ... ... ... ... ... ... ... ... ... ... ... .
17 17-21 .
Heating due to the battery current ... ... ... ... ... ... ... ... ... 18 22 , 23 .
Error in the temperature coefficient of the coils , \amp ; c ... ... .
23 24-26 .
Error in the box thermometer , \amp ; c ... ... ... ... ... ... ... ... ... . .
26 27-29 .
Insufficient immersion ... ... ... ... ... ... ... ... ... ... ... ... ... 28 30 , 31 .
Slowness of platinum thermometers ... ... ... ... ... ... ... ... ... 33 32 , 33 .
Purity of ice , accuracy of barometer , \amp ; c ... ... ... ... ... ... . .
35 34-39 .
Uncertainties connected with the determination of the ct sulphur Point ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... 36 40-45 .
Constancy of platinum thermometers ... ... ... ... ... ... ... ... ... 42 46-49 .
Differences between the thermometers ... ... ... ... ... ... ... . .
4 !
) 50-52 .
Accuracy of the observations ... ... ... ... ... ... ... ... ... ... ... 53 53 .
Improvements suggested in apparatus ... ... ... ... ... ... ... ... .
56 54 .
Further experiments wanted ... ... ... ... ... ... ... ... ... ... ... .
57 o5 .
Acknowledgment of assistance ... ... ... ... ... ... ... ... ... ... ... 57 Preliminary .
S 1 .
In 1895 the Kew Observatory Committee decided , in the words of their annual Report , " to instal platinum thermometers at Kew , and to institute an independent series of experiments into their behaviour .
" The Report adds , " Attention will , in the first instance , be directed more especially to the question of the fixity of the zero and of the fundamental interval .
" In pursuance of this policy the Committee built a special room , furnished with a fume closet , and purchased from the Cambridge Scientific Instrument Company six platinum thermometers and " a Callendar-Griffiths resistance bridge , which was regarded at the time as embodying all the latest improvements .
In the choice of apparatus and the construction of the room , the Committee had the advantage of the advice of Mr. E. H. Griffiths ; while Mr. C. T. Hey cock and Mr. F. H. Neville , as well as Mr. Griffiths^ 4 Dr. C. Chree .
Investigations on kindly visited the Observatory , and illustrated the methods of using the apparatus .
An account of the original installation was given by Mr. Griffiths in ' Nature/ Nov. 14 , 1895 , pp. 39-46 .
As this paper and several more recent papers by Professor Callendar and others have discussed the fundamental facts of platinum thermometry very fully , I have judged it unnecessary to go into such details here .
Thermometers .
S 2 .
The six original thermometers , distinguished as Ki to Kc , were all made from one sample of platinum wire .
In 1896 a seventh thermometer , K ?
, was obtained from the Instrument Company .
It is believed to be of the same sample of platinum wire as the others , but this is not absolutely certain .
Particulars as to the type of the thermometers are given in the following table:\#151 ; Table I. Ther- mometer .
Material of tube .
Length from end to edge of wooden collar .
Length from end to terminals .
Diameter of tube .
Approximate 1 resistance in ohms , answering to fundamental interval .
cm .
cm .
mm. Ki porcelain 33-0 37*5 11 *5 1*0 K , 33 35 5 40*0 11 -5 1 *0 k3 33 26-5 31*0 13-5 1 *0 K4 33 26-5 31*0 13 5 1 *0 K* glass 35 -5 41*5 10-5 1*0 Kfi 33 35 0 41 -0 14-0 2*5 33 34-5 41 -0 8-0 1*0 The resistance at 0 ' C. of K6 is about 6-5 ohms , while the resistance of each of the other thermometers at 0 ' C. is about 2-6 ohms .
Since it came to Kew , K6 has been exposed to no temperature above the steam point ; whilst K3 , Iv4 , K5 , and K7 have not been taken above the sulphur point ( 444''53 C. according to Callendar and Griffiths , under normal pressure ) .
In their early days , Ki and Ko were heated on several occasions to the temperature of melting silver ( approx. 960 ' C. ) .
After a few months ' use , KT began to behave unsatisfactorily .
The Instrument Company reported after inspection that the tube was slightly short , and a new tube was fitted in March , 1896 .
The career of K2 has been chequered .
The first tube broke in melting silver in Platinum Thermometry at Observatory .
5 March , 1896 ; the second tube was found to be broken in August , 1896 , after a silver point experiment .
After repair the thermometer behaved badly , and had to be sent to Cambridge .
A third tube cracked in molten silver in December , 1896 .
The fourth tube lasted until the end of 1898 , when it was broken at Sevres .
On more than one occasion the thermometer had practically to be remade , so that the observations taken with K2 at different stages of its existence are not comparable .
The spiral at present in K2 is believed , however , to be of the same sample of platinum as the original one .
The other five thermometers are not known to have had any misadventures .
The thermometers K2 and K5 were taken by Dr. Harker to Sevres in July , 1897 , and did not return to Kew until the end of 1898 .
No use was made of K5 at Sevres , but some observations were made with K2 up to a temperature of about 600 ' C. S 3 .
In an ordinary platinum thermometer , it is possible for the air inside the tube to become unduly moist , with consequent deterioration of the insulation .
In a glass tube the presence of moisture may be detected by the cloudy appearance when the thermometer is cooled in ice ; in the case of a porcelain tube the only guide is the behaviour of the galvanometer .
If there is no sensible creep in the galvanometer , it is probably best to leave the thermometer alone , even if slight cloudiness is visible .
There is some risk of altering the apparent zero of the thermometer in removing it from the tube , and replacing it after the tube has been dried out .
On the rare occasions when a tube has been dried out , check observations have been taken before and after the process .
The room in which the platinum thermometers are kept is heated night and day when necessary by a gas stove , whose combustion products escape by a flue opening outside .
As the room is naturally a dry one , the risk of moisture has thus been small .
resistance Box .
S 4 .
The box as it originally existed in 1895 had a plug system similar to that of the ordinary Post Office pattern .
The plug holders were of brass .
Towards the end of 1896 it was found that pulling out a plug influenced the tightness of its neighbours .
Early in 1897 the Committee arranged with the authorities of the International Bureau of Weights and Measures that Dr. J. A. Harker should proceed to Sevres , and take part in a comparison of platinum and gas thermometers .
At first it was proposed to take the existing Kew box to Sevres ; but , on hearing of the difficulties experienced with it , the Committee decided that Dr. Harker should examine into their reality before a decision was come to .
Finding that with an ordinary standard of plug tightness sufficiently consistent results were not Dr. C. Chree .
Investigations on attainable , Dr. Harker tried the effect of greater tightness This , however , made matters rapidly worse , and it shortly became obvious that the box in its existing condition was of no further use ; it was accordingly sent for repair to Cambridge .
Having high hopes of a new system of plug holders\#151 ; fusible metal inside Doulton ware\#151 ; the Instrument Company introduced this , at their own expense .
The plug holders in the restored box are supported separately , and the pulling out of a plug has never shown any tendency to influence others .
This is undoubtedly a great improvement ; but I am somewhat doubtful of the expediency of the other changes made when restoring the box .
After some time it was found that taking out and replacing a plug sometimes exerted a very sensible influence on the reading , and from the appearance of the plugs it was suspected that this arose partly from the state of the plug holes .
The fusible metal seems disposed to develop a coating of light-coloured powder , and presumably this affects the plug resistances .
On being applied to , the Instrument Company supplied a simple arrangement for cleaning out the plug holes without undue friction , and it has certainly improved matters .
The accompanying sketch , fig. 1 , shows diagrammatically a vertical section of the original box , perpendicular to its longest dimension .
C represents the coil chamber , AAA a copper tank containing water .
This tank could be heated from below by a gas burner , the flow of gas being determined by a gas regulator , whose mercury bulb was surrounded by the water inside A. The coils hung in air , and their temperature was deduced by means of a mercury thermometer whose bulb was inside C. When altering the plug system , the Instrument Company altered the shape of the water tank and the form of the coil chamber .
Fig. 2 gives a section of the existing form taken in the same way as fig. 1 , Fig- .
1 .
the letters having the same significance .
The changes were presumably dictated by the altered nature of the plug arrangements .
I am disposed to think that the coil chamber in the existing form is not so well protected from external influences as in the earlier form , and I believe that in accepting the readings of the mercury thermometer , as giving the temperature of the coils , there is more risk of error now than formerly .
Platinum Thermometry at Kew Observatory .
7 resistance Coils .
S 5 .
The coil resistances are on the binary scale , H = G/ 2 = F/ 4 = E/ 8 = D/ 16 = C/ 32 = B/ 64 = A/ 128 = Aa/ 256 , with two extra coils , " Cal " and FI\#151 ; or I , as it will be called here the former used merely in calibrating the bridge wire .
The nominal value of H is 5 box units .
The unit is very approximately O'Ol of an ohm , so that the fundamental intervals of all the thermometers except Kg are nearly 100 box units .
This has the advantage that when the reading R0 in ice is subtracted from the observed reading , one has approximately the temperature in degrees Centigrade on the platinum scale .
The coil I is approximately 100 box units , and practically all the fundamental interval determinations , except with Kg , have been referred to it .
The coil A a was added along with the coil Cal in 1897 , when the box was altered .
It has not been used in any of the work now to be described , and the mean box unit has been based throughout on the eight coils H to A. The coils are of platinum silver .
The bridge wire is fully 30 cm .
long , readings to the right of the centre representing a temperature above , and readings to the left a temperature below that answering to the sum of the coils whose plugs are out .
The bridge wire is also of platinum silver , and possesses the same temperature coefficient as the coils .
The bridge-wire scale is divided to 1 mm. , and the vernier reads directly to 0'02 mm. , and allows 0'01 mm. to be estimated .
A difference of 1 cm .
on the bridge-wire reading answers very nearly to 1 box unit , or to a difference of 1 ' in temperature with all the thermometers except Ko .
Thus , in referring to differences of bridge-wire readings I shall usually , for brevity , speak of them as differences of temperature .
Nature of the Observations .
S 6 .
The expediency of using the same coil combinations for all the thermometers except Kg was soon recognised .
Also a point has been made of using two combinations for each temperature , according to the following scheme :\#151 ; Thermometers .
Ki , K2 , K3 , K4 , Kb , K In ice .
f CDF 5 ' ^7 ' ICDGH TAG l AH In steam .
In sulphur vapour .
redfi r ae ICDGHI LAFGH f ACDF lACDGH By using the same coil combinations for different thermometers we may at least hope to detect any sensible relative changes , whilst Dr. C. Chree .
Investigations on by using two coil combinations we get a better idea of the order of accuracy of the observations , and have some check on the constancy of the smaller coil resistances .
As a rule , each fixed point determination depended on at least three observations .
Thus , in ice we might have a balance against CDF , next against CDGH , and finally against CDF again .
By comparing the readings in the first and third observations one can make reasonably sure that the [ thermometer has reached the true temperature .
This is a very necessary precaution , especially in sulphur point observations .
" When only three observations were taken , only half weight was allowed to the first and last , so as to give equal weight to the two combinations of coils .
It was originally intended to take a complete set of " fixed point " observations , ice , steam , and sulphur points , with each thermometer once a month .
The regular observations have been in reality less frequent than was intended , and the intervals between them more irregular .
In addition to the regular observations there have been a variety of special occasions on which ice and steam point determinations have been made , more especially with Ky. This thermometer has been used in comparing occasional high-range mercury thermometers .
Sources of Change or Error .
S 7 .
Before considering the main question , I propose discussing a number of possible sources of change in the readings obtained with platinum thermometers .
An error of invariable amount from a source external to the thermometer is perhaps not immediately germane to this inquiry , but the invariability of error is a very difficult thing to ensure .
Excluding observational error or defects , such as moisture in the tube , which an experienced observer is likely to detect , I would chronicle the following sources of trouble :\#151 ; 1 2 3 4 5 6 7 8 1 .
Change in the resistance of the platinum spiral ; 2 .
Change in the thick platinum wires connecting the spiral to the thermometer terminals , or in the compensating loop inside the tube ; 3 .
Change in the relative resistance of the leads , or in the relative resistance of the proportional arms of the box ; 4 .
Change in the box coils or in the bridge wire ; 5 .
Faulty action of the contact piece ; 6 .
Shift of the bridge centre ; 7 .
Thermo-electric currents ; 8 .
Heating due to the battery current ; Platinum Thermometry at Kew Observatory .
9 9 .
Error in the temperature coefficient of the coils , or differences between the temperature coefficients of different coils ; 10 .
Change of zero or other error in the mercury thermometer inside the coil chamber ; 11 .
Failure of the box mercury thermometer to give the true temperature of the coils and bridge wire ; 12 .
Insufficient immersion of the thermometer ; 13 .
Slowness of platinum thermometers in acquiring the true temperature .
There are other possible sources of error affecting only certain fixed pointobservations , and mostly not peculiar to platinum thermometry , for instance:\#151 ; 14 .
Impurity in the ice used in getting Ii0 , or variability in the method of treating it ; 15 .
Error in the barometer , whether constant or varying , or error in the reduction of its readings ; 16 .
Impurity in the sulphur used , and other uncertainties in the determinations of the boiling point of sulphur ; 17 .
Error in the formula assumed for the variation of the boiling point of sulphur with pressure .
I have seen no adequate consideration of these sources of error in any treatment of platinum thermometry , and in the original equipment obtained for Kew Observatory no provision was made for the detection of most of them ; thus our experience may be useful to others .
Without at least a general idea of their nature , the reader would be unable to judge correctly of the degree of probability attending the conclusions reached , or even of the nature of the evidence on which these conclusions are based .
S 8 .
I propose deferring until the end the discussion of the evidence bearing on the first two items ; only calling attention in the meantime to the fact that without taking a platinum thermometer to pieces it would be very difficult , if not impossible , to distinguish between changes in the resistance of the spiral and changes in the resistance of the thick platinum wires connecting the spiral to the terminals .
The nn alter ability of a platinum thermometer cannot be proved by experiments on an isolated platinum spiral .
The connecting platinum wires inside the tube are by no means of negligible resistance .
They ought to be very approximately equal in resistance to the compensator loops intended to neutralise their variation with temperature .
The resistance of the compensator loops has been measured by Dr. Harker with the following approximate results .
The data answer to a temperature of about 15 ' C. Dr. C. Chree .
Investigations on Table II.\#151 ; Resistance of Compensator Loops .
Thermometer ... ... . .
Ki K2 K3 K4 K5 K6 K7 Resistance ( in ohms ) 020 0'22 0-15 0'15 0'21 0'20 0*22 We may thus infer that the average resistance of the connecting wires is some 7 per cent , of that of the spiral in one of the ordinary thermometers .
Thus a small differential change between these wires and the compensator loop might appreciably affect the thermometer readings .
The Cambridge Instrument Company inform me that the connecting wires and the compensator loop\#151 ; which are of the same thickness of wire\#151 ; are of the same platinum as the spiral and have the same temperature coefficient .
If this were not the case a small difference between the absolute resistances of the connecting wires and the compensator would introduce a source of error fluctuating with the temperature of the wires and compensator .
It would , I think , be desirable to make sure that the process of attaching the thick wires to the fine spiral causes no change in the material close to the junction .
Changes in the Leads or in the Proportional Arms .
S 9 .
In the leads supplied to Kew Observatory one pair has the terminals at both ends marked P , the other pair has its terminals marked C. This is intended to show that the former pair should be attached to the terminals marked P on the thermometer head and to the plugs P in the box , and that the other pair should be attached to the C terminals -on the thermometer head and the plugs marked G in the box .
Supposing this arrangement invariably adhered to , any constant difference between the resistances of the leads should not appreciably influence the apparent resistance of a platinum thermometer .
Supposing , however , that a difference between the resistances of the leads develops itself , then if the leads are used in the way indicated above , the effect on the readings is the same as if the platinum thermometer itself changed .
Loosening and tightening the leads had been found early in the investigations to be a possible cause of slight variations in the reading \#151 ; owing presumably to the varying tightness\#151 ; and it may allow the terminals to blacken during the sulphur point experiments .
It was thus at first thought undesirable to touch the lead connections once the thermometer was attached , especially as it was supposed that if error were at all likely to arise in so obvious a way , provision would have been made for detecting and eliminating it .
Thus the matter did not receive consideration until 1897 , when the erratic nature of some of the results drew attention to the fact that the leads were becoming worn near the ends .
Platinum Thermometry at Kew Observatory .
11 They were repaired by the Instrument Company , and on their return in September , 1897 , a large discontinuity was observed in the values of R0 in the platinum thermometers .
I then found that interchanging the leads caused a difference of fully 1 ' in the reading .
This seeming unsatisfactory , the leads were readjusted by the Instrument Company , and we assured ourselves of their approximate equality , without , however , attempting to measure the exact difference .
As the leads remained satisfactory to the eye , no more attention was given to the matter until the autumn of 1898 , when a conspicuous drift appeared in the values of Ro .
Experiment then showed that the difference between the leads was liable to large fluctuations ; accordingly new leads were got from the Instrument Company in December , 1898 .
These had been adjusted with great care , and on their arrival at Kew they were found to be very nearly though not exactly equal .
Table III gives a brief synopsis of our experience with the leads .
The quantity tabulated is the excess of the box reading obtained\#151 ; the thermometer being in ice , steam , or sulphur vapour\#151 ; when the CC leads were connected to the thermometer over the reading when the PP leads were so connected .
Table III.\#151 ; Effect of Interchanging Leads .
Date .
Number of Thermometers Observed differences .
experi- ments .
riised .
greatest .
least .
mean .
1897 .
Sept. 21 1 Ki 1-029 [ Leads adjusted .
] 1898 .
Oct. 18 6 0'"432 0 " 391 0 -414 Oct. ] 9 , 20 8 0 -473 0-438 0-453 Nov. 7 , 8 , 9 , 10 , 11 6 KiK ; jK4K7 0-550 0-494 0-517 Nov. 26 1 K7 \#151 ; 0-257 Nov. 28 , 29 , Dec. 8 ... .
4 K7 0-376 0-366 0-372 [ New leads .
] Dec. 8 , 12 , 13 4 KjKgK , 0-026 0 -018 0-023 1899 .
Jan. 24 , 25 4 KjKgZgK ; 0 -033 0-011 0*020 May 10 , 17 2 k , k4 0-026 0 013 0-020 June 27 t. t 1 K2 - 1 0*032 July 7 , 10 7 KjKs 0 -062 0 045 0*054 | Sept. 14 , 15 , 18 , 19 , 20. .
8 K1K3K4K5KgK7 0-044 0-028 0*036 Each experiment consisted of at least two readings with each arrangement of the leads .
The resistance of one pair of the new leads was recently found by Dr. Harker to be approximately 0'07 ohm at atmospheric ternDr .
C. Chree .
Investigations on perature , i.e. , about 7 per cent , of the fundamental interval of one of the ordinary thermometers .
Thus a change of one-thousandth part in the resistance of one of the leads would answer to an alteration of about 0014 in the quantity tabulated in Table III , or to an apparent shift of 0o,007 in the zero of an ordinary thermometer .
There are various suggestive features in the table .
Between October 18 and November 11 , 1898 , the difference between two leads , immaculate to the eye , altered steadily in one direction by an amount equivalent to a change of 0o,05 in the zero of a platinum thermometer .
Then , without apparent cause , there occurred a sudden alteration in the opposite direction , equivalent to a change of 0'T2 in the zero .
After this very erratic behaviour the leads seem to have remained practically constant for ten days .
Next we come to absolutely new leads of the best construction .
No one , I think , can question the reality of the differences between the results of December , 1898 , July , 1899 , and September , 1899 .
But with the occurrence of differences of the size shown during 1899 , we must anticipate errors of the order of at least 0''01 or 0o,02 in the absolute readings , unless adequate provision is made for eliminating this source of uncertainty .
Changes in the leads have unquestionably been the principal cause of the apparent changes in the zeros of the platinum thermometers during the investigation .
These changes have added considerably to the difficulty of working up the results .
The difference between the proportional arms was recently found by Dr. Harker to be less than 1/ 6000 of either resistance .
Doubtless the resistances were originally made as nearly equal as possible , so that there is a strong presumption that if any differential change has occurred it has been small .
There is , however , unfortunately no direct evidence bearing on this point .
S 10 .
It is only proper to remark that from some points of view slow , regular changes in the relative resistances either of the leads or of the proportional arms are not of primary importance .
Such changes , at least when small , are equivalent , the one to the addition of a constant quantity to the observed resistances , the other to the multiplication by a constant factor , so long as the measurements dealt with cover only a short time during which no sensible variation occurs .
Supposing sufficiently numerous observations made of the resistances R,0 and Ri in ice and steam , it is clear that the platinum temperature pt , given by pt \#151 ; 100 ( R \#151 ; R0 ) / ( Ri \#151 ; Ro ) , would remain practically unaffected .
But , on the other hand , the necessity of frequent zero point observations is one of the drawbacks most frequently dwelt upon in the case Platinum Thermometry at Kew Observatory .
13 .of mercury thermometers , and it is only fair to recognise that they do not in this respect necessarily suffer by comparison with platinum thermometers .
Coil Changes .
S11 .
The box was originally calibrated at Cambridge by Mr. E. H. Griffiths , who referred everything to a mean box unit based on the eight coils H to A. Since the box came to Kew Observatory it has been thrice calibrated , with the aid of a convenient apparatus designed by Dr. Harker .
The observations were made on the first of these occasions by Dr. Harker ; on the other two occasions mainly by Mr. W. Hugo , Senior Assistant at the Observatory .
I took a certain number of check observations during the two last calibrations , and made all the calculations necessary to construct the correction tables based on these and on Dr. Harker 's calibration .
In doing so , I followed Mr. Griffiths ' procedure , except in some minor details .
The results of the four calibrations were as follows :\#151 ; Table IV.\#151 ; Coil Values in Terms of Mean Box Unit .
Coil .
Nominal value .
Calibration I ( Sept. 1895 ) .
II ( May , 1897 ) .
III ( July , 1897 ) .
IV ( Mar.-Apr .
, 1899 ) .
A. 640 639 -941 639 -973 639 -980 639-975 B 320 320 -169 320 -093 320 -090 320 -090 C 160 160 -062 160 -023 160 -025 160-028 D 80 80-010 80 -001 79-991 79 -991 E 40 39 -968 40 -037 40 -028 40-023 F 20 19 -963 20-111 20 -115 20-115 a 10 9-945 9 -760 9-766 9-774 H 5 4-942 5-002 5-005 | 5-004 i 100 99 -957 99-989 99 -975 99 -426 Mean 1 cm .
bridge wire 0-995 0-993 1 0-994 0-995 The considerable differences between the results of Calibrations I and II are probably mainly due to changes attending the reorganisation of the box in the spring of 1897 .
Coil I broke at one end , and had to be resoldered just before the last calibration .
It is improbable that the algebraic sum of the resistances of the eight coils H to A was unaffected by the box changes in 1897 , and thus the mean box units in Calibrations I and II were almost certainly slightly different .
The data obtained with the thermometers , as will be seen later , suggest a slight increase at this time in the mean box unit ; and this is , at least , consistent with the apparent change in the bridge-wire resistance .
Dr. C. Chree .
Investigations on We should rather expect the resistance of the bridge wire to increase , especially near its centre , through constant rubbing by the contact piece , and the figures obtained in Calibrations III and IY somewhat favour this view .
Too much significance ought not , however , to be attached to the small differences apparent .
Some of the differences in the coil values in Calibrations II , III , and IV may be experimental errors ; but the changes shown in the case of , at least , E and G are , I think , too large to be accounted for in this way .
S 12 .
Perhaps the clearest evidence of the reality of coil variations is that afforded by an examination of the thermometric results obtained with the different coil combinations .
It would occupy too much space to go into details , so I merely record in Table V the mean differences between the values of E0 , Ri , and Rs ( resistance in sulphur vapour ) obtained with the two coil combinations used during different specified epochs .
The unit in the table is O'Ol mm. of bridge wire , answering approximately to 0o-001 C. with the ordinary thermometers .
In the final means equal weight is allowed to each observation , so that some thermometers exert more influence than others .
During some of the epochs\#151 ; especially 1895 and the first part of 1899\#151 ; the data with any one thermometer were very scanty .
Again , it must be remembered that the part of the bridge wire at which readings are taken differs according to the thermometer used , and also varies for any one thermometer according to the temperature of the coils , the position of the bridge centre , and the difference between the leads .
It is also different in the ice , steam , and sulphur point observations , fluctuating considerably in the latter two cases with the barometric pressure .
Thus the fluctuations in the table amongst the results for a common epoch are due to many causes .
S 13 .
It is , I think , most instructive to start with Calibration III made in July , 1897 .
During the rest of that year the results from the two coil combinations show almost perfect agreement ; in 1898 the results drift apart , and the drift is accentuated in 1899 prior to Calibration IY .
Again , for some months after that calibration there is an excellent agreement , though a tendency to drift soon manifests itself .
Calibration II seems less successful , but it was made by an observer different from the one who took the readings on which the results in Table Y are based .
A different standard of plug tightness does not influence all plugs alike , and in my experience the personal equation in this matter requires to be reckoned with .
The data from March 12 to 19 , 1897 , were so outstanding that they are given separately .
The exposed parts of the box had , I believe , been cleaned shortly before , and conceivably one of the coil supports may have got a knock .
There was , however , no suspicion of this at the time .
This fact emphasises the necessity of a constant outlook for possible changes .
Table Y. Mean excess of readings with combination C D F ( I ) bination C D G H ( I ) .
over reading !
3 with com- Mean for all , Mean excess of reading with combination A E over reading with combination Mean A a 1 In ice .
In steam .
both ice and steam .
A F G H in sulphu r vapour .
for all .
o p-i H 3 Ki- k2 .
k3 .
k4 .
w 1 !
1 * i K , .
k2 .
k3 .
k4 .
k5 .
k6 .
k7 .
Ki .
K\#187 ; .
K3 .
k4 .
k5 .
k7 .
1895 i - 9 -28 -46 . . . .
-18 - 8 + 22 . .
+ 28 . .
-10 . .
0 -18 -22 . .
, , -13 189(3 i -11 -12 -14 - 9 -10 -26 -13 - 7 -12 - 7 -11 -12 -20 -11 + 14 - 3 + 11 + 16 + 5 . .
+ 8 1897 Jan.- i Feb. . .
-23 -10 -16 + 9 - 8 . .
- 2 . . . . . .
- 2 - 6 1897 I Mar. 12-19 .
+ 41 + 4(3 + 62 + 55 + 60 + 46 + 54 + 62 + 41 + 44 + 42 . .
+ 54 + 50 1897 II June .
+ 22 + 20 + 15 + 5 + 12 + 8 + 16 + 18 + 24 0 + 7 + 10 + 6 + 13 + 28 . . . . . .
+ 36 + 32 1897 Ill + 3 . .
+ 3 + 4 . .
0 + 4 . .
- 3 + 3 . .
- 3 -10 - 1 + 4 . .
0 - 8 . .
- 1 - 1 1898 III - 5 . .
-10 -14 . .
-17 - 3 . .
-13 -13 . .
-30 - 8 -12 + 1 . .
-25 -26 .
, -19 : -17 1899 III -10 . .
-15 -29 -34 -48 -14 . .
-18 -32 -19 -51 -50 -28 ~ 8 , . .
-48 -45 -10 - 6 -23 1899 May\#151 ; IV July .
+ 3 - 1 + 8 + 2 + 8 1 - 3 + 6 + 1 + 14 | + 7 + 4 - 6 + 7 + 4 + 10j - 4 \#151 ; 22 + 1 + 21 + 4j + 2 ^2 c\gt ; ^ * I \amp ; O Co ( -U CJl Dr. C. Chree .
Investigations on A complete calibration is a laborious process , and its frequent repetition would add seriously to the work entailed by platinum thermometry .
If , however , the results in Table V are normal , it would appear that frequent calibrations can hardly be avoided in the case of physical work of the highest accuracy .
Faulty Jction of the Contact Piece .
S 14 .
Parallel to the real bridge wire in the Kew resistance box runs an exactly similar wire connected to the galvanometer , and the contact piece works by pushing down a short cross wire so as to span the interval between the wires .
In January , 1898 , it was noticed that mere moving and resetting the contact piece might alter the reading , and I found that the pressure to which the cross wire was exposed when contact was made\#151 ; always acting on the same part of its surface \#151 ; had cut a groove in it .
The Instrument Company put this to rights , but the phenomenon had repeated itself by January , 1899 , though to a smaller extent .
On that occasion the Instrument Company made an alteration which , it is hoped , will prevent the recurrence of this trouble .
It is difficult to keep the two parallel wires equally tight and exactly at the same horizontal level ; thus the cross wire may bear unduly heavily on one wire before it makes good contact with the other .
This defect would be of little consequence in an open scale bridge wire ; but in the Kew box it produced , when at its worst , uncertainties of the order 0''04 , or even 0''05 , in individual readings .
Of course this merely tended to introduce irregularity in the readings , and supposing a number of observations taken , could hardly simulate a change in a thermometer .
Shift of the Bridge Centre .
S 15 .
By the bridge centre I mean the vernier reading when a balance is made with all the plugs in their holes , the platinum thermometer and compensator resistances being cut out by short-circuiting straps .
The original departure 0'006 in this centre from zero on the scale was given as a fixed correction in the first calibration table , and as no provision existed for determining the centre , I did not for some time properly appreciate the situation .
My attention was first roused by a sudden apparent discontinuity in the values of B0 in the spring of 1897 , for which the only apparent cause was a cleaning up of the box .
The Instrument Company then supplied two short-circuiting straps , to be put across the CC and PP box terminals , and since then determinations of the centre have been made at the beginning of each observation day , and usually at intervals throughout it .
Soon it became apparent that during a day 's observations the bridge centre is apt to drift towards the minus side of the scale ; the Platinum Thermometry at Kew Observatory .
17 same side as it goes to when the plugs are dirty and have an increased resistance .
The amount of the drift is very variable .
One day it may be negligible , while the next day it may be as much as 0'*04 ; it is occasionally more than this .
The phenomenon is probably partly due to the fact that frequent pulling out and putting in of the plugs seems inclined to produce the grey powder already referred to .
It may also arise from the observer 's standard of plug tightness falling off as he becomes tired , or from asymmetric heating of the box through the proximity of the observer 's person to the minus side of the scale .
There are probably various influences at work , some of which may be peculiar to the particular box .
One cannot be perpetually taking bridge centres , so that even the more recent observations are exposed to some uncertainty from this source .
The practice of regularly taking bridge centres was not introduced until 1897 , and it is of course possible that there was little or no occasion , for it in the original box with brass plug holes .
Thermo-electric Currents .
S 16 .
The principal seat of thermo-electric currents seems to be the junctions at the head of the thermometer ; and it is desirable to shield the head as much as possible from heated air or vapour .
When a Griffiths key is used these currents are not necessarily a source of error , but they tend to increase the difficulty of taking readings .
Originally it was supposed that the use of a Griffiths key rendered further care unnecessary , and no commutator was used until after the restoration of the box in 1897 .
As the Instrument Company then transferred certain terminals to the key-board , it is possible that the original box suffered but little if at all from the source of trouble now to be described .
When the commutator was introduced the following phenomena were observed .
When the day 's observations began the readings with the current d and r ( direct and reversed ) were usually about the same .
After a few minutes ' observations the difference between the d and r readings began to increase , and nearly always in one direction .
After a little the difference usually assumed a fairly constant value , but considerable fluctuations might occur , especially if the temperature of the room was rendered unsteady .
On the advice of Mr. W. N. Shaw the thermo-electric key was enclosed in a padded box .
This has decidedly diminished the evil , but it still appears expedient to take readings with the current both ways .
Table VI shows the state of matters typical before and after the introduction of the protecting box .
When the thermo-electric effect is not eliminated the error in the reading is ( cl r)/ 2 .
Supposing d \#151 ; rto remain constant through the ice , steam , and VOL. LXVII .
C Dr. C. Chree .
Investigations on Table VI .
Without protecting box .
With protecting box .
Thermometer .
Hour .
Difference r\#151 ; d. Thermometer .
Hour .
Difference r \#151 ; d. h. m. h. m. 11 45 0 '000 11 o 11 54 0-064 -"-I 11 J 11 35 O *000 0*003 12 12 0 -094 12 20 0*017 12 39 0-106 K , 12 50 0*016 ^3* 2 56 0-086 2 31 0*020 3 30 0-105 k3 2 50 0*014 ^4 3 41 0-107 3 33 0*018 4 20 0-121 k4 3 50 0*027 3 59 0*022 4 26 0*019 sulphur point observations with a particular thermometer , the R0 , Ri , and R* would be wrong to the same amount , so that the fundamental interval and ptt ( the value of pt answering to the boiling point of sulphur ) would be unaffected .
In practice , however , we cannot anticipate so favourable a contingency .
After some experiments , I found that I could at pleasure change the amount , and even alter the sign of the difference by heating with the finger or cooling by an air-blast one of the terminals of the Griffiths key .
Heating due to the Battery Current .
S 17 .
In the Kew apparatus there is an option of two resistances , viz. , 100 and 20 ohms , in the battery circuit and our original instructions were to use the 100 ohms when observing with an ordinary thermometer in ice and steam , and the 20 ohms when observing at the sulphur point .
The object , doubtless , was to ensure sufficient galvanometer sensitiveness at high temperatures .
In an early sulphur experiment I was surprised to find that the reading was about 0'-07 higher with the 20 ohms in use than with the 100 .
There being only a single dry cell in use , I had not anticipated a sensible difference .
Supposing the celbto remain constant , the heating effect would of course remain the same for a given thermometer at any one fixed point , and so would not influence the results immediately in view .
It seemed , however , inexpedient to trust to this , and we have accordingly employed the 100 ohm resistance in all the regular observaPlatinum Thermometry at Kew Observatory .
19 tions , merely using the 20 ohms occasionally to furnish data from which to calculate the heating effect under the normal conditions .
To understand how this is done we must glance briefly at the theory .
Let E be the E.M.F. , R ' the internal resistance of the battery , R the rest of the resistance in the battery circuit .
Let be the resistance in either of the proportional arms , r the resistance in the bridge arm containing the platinum thermometer , whose spiral has a resistance p. Then if i denote the current in the spiral , H the heat given to it in unit time , we have H = i2p \#151 ; EVi2/ \gt ; -7-[(R + R ' ) ( r + ri ) + 2 ?
Ti]2 ( 1 ) .
The heating of course is gradual , and theoretically it might be possible , by rapid , skilful manipulation of the tapping key , to obtain a balance before there is a sensible effect .
In practice , however , this is hardly possible in work of the highest accuracy .
Only by the remotest chance does one hit the balance at the first attempt , and , as a rule , the key must be put down a good many times .
Also , unless the key is held down a sensible time , a small absence of balance may be overlooked .
The weaker however the current , the longer the time before sensible heating exists , and with the 100 ohms in the circuit it seems possible to get a fair balance before the heating effect is appreciable .
Thus in comparing platinum and mercury thermometers at high temperatures , where accuracy of the order ( T'01 C. is usually much above what is necessary , the bridge has been regarded as balanced when no movement appears in the galvanometer on depressing the key .
In the fixed point observations , on the other hand , the resistances have been adjusted until no movement appears on releasing the key .
Under these circumstances the current has exercised its full heating effect .
The platinum spiral is presumably heated to such small extent above the surrounding temperature as is required for its gain of heat from the current to balance the loss by radiation , conduction , \amp ; c. We may pretty safely assume that this excess of temperature is proportional to the heat given to the wire , but it must also depend on the specific heat of the platinum , on its radiating and conducting power , and also conceivably on the shape , dimensions , and material of the enclosing tube .
I understand that several authorities* have proposed to keep i2p constant in the spiral by suitably altering the battery resistance .
This , however , as the above reasoning shows , cannot secure constancy in the temperature excess of the spiral at all temperatures .
S 18 .
The application of the formula ( 1 ) presents difficulties .
We know\#151 ; at least , very approximately\#151 ; the resistance rx of the proportional arms , and the resistance p of the spiral , at any " fixed point " temperature , but it is not customary to measure directly the resist* For instance , Waidner and Mallory , * Phil. Mag. , ' July , 1899 , p. 14 .
C 2 Dr. C. Chree .
Investigations on ance r. It exceeds p by an amount equal to the resistance of the leads , including the thick platinum wires inside the thermometer tube .
The mean temperature of these latter wires depends partly on the temperature of the spiral , partly on the length of tube immersed , and partly on the temperature of the external air .
Of course we can measure r on any given occasion , but its fluctuations would be considerable under normal conditions .
My object being rather to get a fairly exact general idea of what to expect under the varied conditions of normal use , than to measure the phenomena with extreme exactness under an arbitrary set of conditions , I have accepted for r a rough approximation , good enough for my special purpose .
The internal resistance of a dry cell varies very much according to its freshness .
I have thus made two calculations , in which this resistance , IT , is taken as 0 and as 5 ohms respectively .
The latter figure actually applies to the later experiments on the relative heating in ice , steam , and sulphur vapour ; while the former answers sufficiently to the case of a new cell .
The E.M.F. of a dry cell falls off in time , though not to a very large extent , and for our present purpose such a change does not concern us .
We do not need to know the value of H , as given by ( 1 ) , in absolute measure , but only the relative values of H with varying p and r , but with constant E and r\ .
In the following calculations I have supposed = 5\#151 ; the proportional arms having each a resistance of 5 ohms\#151 ; and for an ordinary thermometer p = 2'6 in ice , 3*6 in steam , and 6'8 in sulphur , while r - p \#151 ; 04 ; for we have p = 6 '45 in ice and 8'95 in steam .
The calculated values of ( H/ E'-Vi2 ) x 107 are given in Table VII .
Table VII.\#151 ; Values of ( H/ E2^2 ) x 107 .
Fixed point .
E ' .
Ordinary thermometers .
k6 .
E = 100 60 40 30 20 Difference 20 and 100 E = 100 20 Difference 20 and 100 fo 38 100 212 357 720 682 Ice . . . .
15 34 \#151 ; \#151 ; \#151 ; 491 457 37 484 447 fo 41 107 225 375 744 703 \#151 ; _ _ \#151 ; Steam* .
15 37 92 182 286 513 476 35 437 402 fO 41 \#151 ; \#151 ; \#151 ; 681 640 \#151 ; !
Sulphur L5 37 \#151 ; \#151 ; \#151 ; - 478 441 \#151 ; \#151 ; " !
The most striking features in the table are the comparative smallness of the differences between the supplies of heat at the three fixed Platinum Thermometry at Kew Observatory .
21 points , with a given value of It ' , and the relatively large effect of an addition of 5 ohms to the battery circuit when the rest of the resistance is only 20 ohms .
No very serious error would arise from treating the heat supplied as independent of the temperature , at least up to 445 ' C. , so long as there is a large resistance of the order of 100 ohms in the battery circuit ; and under these conditions a moderate change in the resistance of the battery itself is not of much consequence .
S 19 .
At any one fixed point , as already stated , we may reasonably expect the rise of temperature of the spiral to be proportional to the supply of heat .
It is desirable , however , to check this conclusion by experiment .
I thus compare in Table VIII the above theoretical results with some experiments made with thermometers K2 and K5 at the steam point in June , 1897 .
In these the additional resistance in the battery circuit\#151 ; beyond that of the battery itself\#151 ; was given several values intermediate between 20 and 100 ohms .
Assuming Heating effect/ heat supplied = C , a constant , I determined C by equating the observed and calculated values of the difference of the heating effects with the two extreme additional resistances 20 and 100 ohms .
This quantity is denoted by H2o - Hioo , and a similar notation is employed for the other similar difference effects .
The differences in the table are really lengths of bridge wire , but for practical purposes they may be treated as temperatures .
Table VIII.\#151 ; Heating Effects of Battery Current at Steam Point .
R ' .
h20-h100 .
H30"H 100*| H40\#151 ; H100 .
1^60"-^100- H\oo* H60 .
H.o- H30 h20 .
^ f Calculated ^ Observed ... 0 0-165 0-078 0-043 0-015 o-oio 0*025 0-053 0-088 0-175 5 0*086 0-050 0*019 0-013 0*032 0*063 0-099 0-178 ... o-ioo 0*047 0*019 K f Calculated^ ^.Observed ... 0 0*185 0-088 0*048 0-017 o-ou 0*028 0*059 0-099 0-196 5 0*097 0*057 0*022 0-014 0*036 0*071 0-111 0-199 ... " 0*097 0*061 0-026 The resistance of the battery itself was not determined on either occasion ; but from the large size of the heating effects compared to those of later experiments with an older cell , I am inclined to suppose that the resistance was nearer 0 than 5 ohms .
The values calculated with R ' = 5 , however , unquestionably present the closest accordance with experiment .
In fact , taking into account the difficulty of the experiments , the agreement in this case is altogether too good , and must be largely a matter of luck .
However this may be , there is , I think , every reason to suppose that the values deduced for the heating 22 Dr. C. Chree .
Investigations on effects Him , \amp ; c. , with the additional battery resistances 100 , \amp ; e. , are a very fair approximation to the truth .
There seems little doubt that with a fresh dry cell the reading at the steam point is raised fully the .
hundredth of a degree , even with 100 ohms in the battery circuit .
If 20 ohms only were inserted , the heating effect would amount to nearly two-tenths of a degree .
In this latter event , an increase of 5 ohms in the battery resistance would lower the observed value of Ri by about 0'-06 C. S 20 .
We have next to consider the relative values of the heating effects at the three fixed points .
These have been determined by a series of special determinations of the values of H20 - Hi00 .
During the majority of the experiments the battery resistance was approximately constant , and not far from 5 ohms .
The following table gives the observed values of the ratios ( H2q - Hioq at steam point)/ ( H20 - Hi00 at ice point ) and ( H20 - H10o at sulphur point)/ ( H20 - H100 at ice point ) for the several thermometers , as well as the values calculated from Table VII for the ratios of the heat supplies .
Table IX.\#151 ; Differential Heating Effects at Ice , Steam , and Sulphur Points ( found experimentally ) .
Kp K2 .
Kg .
K4 .
k5 .
k7 .
Mean Kx to K7 .
k8 .
Steam/ ice 0-77 0-77 0*74 0-75 0-67 6-75 0*74 0-54 S9 ) ) . . . . .
j Ratio of heat supplies ... .
1 _*jdbd II II Ox O 1-03 1-04 0-90 Sulphur/ ice .
... | 0-33 0-26 0 17 0-40 0 -42 0 -29 0-31 99 99 . . . .
Ratio of heat supplies. . . .
f W = 0 ' ' ' t K ' = 5 1 0-94 0-97 The table is based on eighteen experiments comparing the ( steam/ ice ) ratios , and eight experiments comparing the ( sulphur/ ice ) ratios .
In most cases the several steam/ ice observations with the same thermometer showed a remarkably good agreement , but the results with K5 and K7 were less harmonious than the others .
The sulphur/ ice observations were so few that little weight attaches to the results for the individual thermometers j but the final mean should not be much in error .
The highest value of the sulphur/ ice ratio observed was 0-42 in the single experiment made with K5 .
A considerable number of heating experiments have been made in sulphur , but I have employed in Table IX\#151 ; as in the case of steam\#151 ; only those which were made on the same day Platinum Thermometry at Observatory .
23 as a corresponding experiment with the same thermometer in ice .
This is necessary on account of possible changes in the resistance of the dry cell .
At the sulphur point the heating effect is only about one-third of what it would he if it depended only on the value of i2p in the spiral , and even at the steam point the heating is less than three-quarters of what it would be on this erroneous hypothesis .
S 21 .
If the heating effect due to a given were the same at all temperatures , Tables YII and VIII would justify the conclusion that with 100 ohms in the battery circuit all that is to be feared is a trifling fall , of the order 0''003 , in the apparent zero R0 as the cell deteriorates .
But , according to Tables VIII and IX , the true conclusion is that even with 100 ohms in circuit the heating effect is likely to diminish the fundamental interval by several thousandths , and the ice-sulphur interval by not much less than a hundredth of a degree .
The use of 20 instead of 100 ohms in sulphur point observations would largely increase the uncertainty .
The changes in Rs - R0 and Ri - R0 , due to varying heating effect , are in the same direction , so that the effect on is relatively small , and , in view of the numerous uncertainties existent , I have not thought it worth while to attempt a " reduction to an infinitely small current .
" It is clear , however , that unless the current can be materially reduced by the employment of a more sensitive galvanometer , the necessity of providing for this reduction must be kept in view in all work of the highest accuracy .
Before quitting the subject , I would remark that the diminution of the heating effect of a given i2p as the temperature rises is only what we should expect from the known increase with temperature of both specific heat and radiating power .
It is possible that heating effect experiments with thermometers of other metals than platinum , or with spirals of varying gauge of wire , might prove a useful method for investigating radiating power .
Error in the Temperature Coefficient of the Coils or Differences between the Temperature Coefficients of Different Coils .
S 22 .
The standard temperature selected by Mr. Griffiths was 20u C. , and he found for the relation between the resistance of a sample of the coil wire at temperature r , and its resistance the standard temperature , rT \#151 ; r-io [ 1 + 0-00026 ( t - 20 ) ] .
( 2 ) .
All temperature corrections have been deduced by ( 2 ) since the box came to Kew Observatory .
Suppose after the calibration corrections are applied that the observed box reading makes p the resistance of a certain platinum Dr. C. Chree .
Investigations on thermometer when the coils and bridge wire are all at temperature then , 1 eferred to the mean box unit at 20 C. , the true resistance is by ( 2 ) P [ 1+0-00026 ( t-20 ) ] , and thus the proper temperature correction is 0-00026 ( t - 20)p .
Now suppose lit the true resistance at of a platinum thermometer whose true resistance at 0 ' is R0 , and whose fundamental interval is I. Then the platinum temperature is pt= 100(Rt-R0)/ I ... ... ... ... ... ( 3 ) , and corresponding to any correction ( or error ) ARf , arising from a cause other than errors in R0 or I , we have for the correction ( or error ) in pt the formula Apt = ARE *(100/ I ) ... ... ... ... ... . .
( 4). .
Supposing ( 2 ) correct , and the box at temperature r , we must have in order to balance the bridge R\#171 ; = P [ 1 + 0-00026 ( t - 20 ) ] ... ... ... ( 5 ) .
If then Apt is the correction to pt arising from our application of a temperature correction to the box readings , we have Apt = ( 100/ I)px 0-00026 ( t-20 ) ... ... ... ... ... . .
( 6 ) , where p , like I , is measured in mean box units .
For one of the ordinary Kew thermometers the following are sufficiently near values for purposes of illustration :\#151 ; Ice point .
Steam point .
Sulphur point .
P =258 358 678 Corresponding to these representative values , with I = 100 , we have Ice point .
Steam point .
Sulphur point .
Apt = 0-067 ( r-20 ) 0-093 ( r-20 ) 0-176 ( t-20 ) Using Callendar 's formula , t-pt = S[(*/ 100)2-*/ 100 ] ... ... ... ... .
( 7 ) , we have for the small increment At on the air scale corresponding to a small increment Apt on the platinum scale At = Apt -I- [ 1 \#151 ; 10"4 8(2t - 100 ) ] ... .
( 8 ) .
Assuming for the sulphur point on the air scale Callendar and Griffiths ' Platinum Thermometry at Kew Observatory .
25 value , tg = 444'-53 , then in the Kew thermometers , as will he seen later , S = 1 *5 nearly , and we have approximately At the ice point A t = A ptx 0-985 = 0066 ( r - 20 ) } 'j " steam " At = AptfxT015 = 0-094(r-20 ) , \gt ; ... ( 9 ) .
" sulphur " A t = A^x 1-134 = 0-200 ( r-20 ) J S 23 .
Supposing that we aim at an accuracy of 0'-001 C. , and are able to keep our coils within 1 ' C. of the standard temperature , then it will suffice to know our temperature coefficient correctly to 1 part in 66 , to 1 part in 94 , or to 1 part in 200 , according as the " fixed point " concerned is the ice , steam , or sulphur point .
Accuracy to 0 -001 C. at the boiling point is not a very extravagant aim , and it thus appears desirable to know the temperature coefficient correctly within 1 per cent , unless we can keep nearer than 1 ' to the standard temperature .
From correspondence with Mr. Griffiths , I found that he did not anticipate so high a degree of accuracy as this in his determination of the temperature coefficient , and as a matter of fact two fairly complete experiments by Dr. Harker have given values from 0-00025 to 0-000245 for combinations of several of the actual box coils .
The allowance of a departure of 1 ' C. from the standard temperature is a narrow one , except under specially favourable conditions .
A good thermo-regulator , no doubt , can control the gas supply with considerable accuracy , even under unfavourable conditions ; but what the thermo-regulator controls is its own temperature .
That temperature is probably in general close to the mean temperature of the immediately surrounding water .
But supposing , as may happen in the platinum-thermometer room at Kew , that the air temperature is only 14 ' one day and 19 the next , then the temperature of the coil chamber will be much higher on the second day than on the first , supposing the thermoregulator to have been left untouched .
One cannot tell in advance what is going to happen during the night , and if one finds in the morning the box thermometer indicating 17 ' and attempts to bring it up to 20 , then either a long time is wasted in waiting during a very gradual rise , or else one is very uncertain as to the uniformity of the temperature in the coil chamber .
Again , in summer the room may be at 22 ' or 23 ' , or more , and the only simple way of keeping the coil chamber at 20'\#151 ; viz. , constant supplies of cold water in the bath\#151 ; would probably introduce uncertainties greater than those it removes .
On the whole , experience at the Observatory suggests that it is best to keep the box nearly at the temperature of the room , so that seasonal* variations of t between 17'"5 and 22'"5 have been common .
Supposing * In the course of a single day 's experiments the temperature changes were usually only a few tenths of a degree .
Dr , C. Chree .
Investigations on t-20 = 2''5 , an error of 5 per cent , in the temperature coefficient would introduce the following errors in t:\#151 ; Near 0 ' C. Near 100 ' C. Near 445 ' C. 0-008 ' 0-012 0-025 The above remarks suppose that we are concerned , as in the Kew experiments , with the absolute value of the platinum thermometer resistances , and have to do with observations taken on different days .
In ordinary thermometric work uncertainties in the temperature coefficient would be much less serious , supposing one to determine the fundamental interval daily , and to combine together only such observations as are taken on the same day .
Under such circumstances it would not matter how far removed from 20 ' the temperature was , or how much in error the temperature coefficient was , so long as the temperature of the coils was strictly constant .
Unless , however , the room containing the platinum thermometers is kept at a constant temperature artificially , or is exceptionally situated , its diurnal range of temperature will seldom be inappreciable .
Another source of uncertainty is the possibility of differences between the temperature coefficients of the different box coils , or between those of the two proportional arms .
Such differences have not been observed at Kew , but there has been hardly any direct experiment , and the comparison of the results from two coil combinations , such as CDF and CDGH , in which only the small coils are different , does not afford a delicate test .
As a difference between the temperature coefficients of the proportional arms was suggested by the Cambridge Instrument Company as a possible explanation of some phenomena ( due , however , to some other cause ) , it is presumably a contingency which should not be disregarded .
Error in the temperature coefficients , it should be noticed , would only introduce irregularity into the results of the Kew experiments .
It might conceivably introduce an apparent seasonal variation in Ro or in the fundamental interval , but could hardly simulate a true secular change .
Change of Zero or other Error in the Mercury inside the Coil Chamber .
S 24 .
An error of 0'T C. in the estimated temperature of the coils affects the calculated temperature in the case of one of the ordinary Kew platinum thermometers by 0'"007 at 0 ' C.,4)'-009 at 100 C. , and 0'-020 at 445 ' C. Thus for accuracy of the order O 'OOl at 100 ' C. , or 0 " 002 at 445 C. , the temperature of the coils must be known to 0 " 01 C. An ordinary Platinum Thermometry at Observatory .
27 uncalibrated mercury thermometer , even with an open scale , cannot be relied on to give temperature differences to this degree of accuracy under the most favourable conditions .
The thermometer in use in the Kew box is a good calibrated one , and errors in its graduation are unlikely to exceed 0'*01 .
It is , however , subdivided only to fifths of a degree , and individual readings cannot claim an accuracy of 0''01 .
Fortunately , errors in reading merely introduce irregularities in the results , and so are not of fundamental importance in the present inquiry .
A more subtle source of trouble is the secular change of zero , normal to mercury thermometers .
A rise of 0'T in the zero\#151 ; which is not very much in excess of what actually occurred during the four years occupied by the experiments\#151 ; would , in fact , exactly simulate a rise of 0o-007 in the value of Kq .
In the ordinary use of platinum thermometers , at least when there are moderately frequent observations of B0 , the circumstances are different .
The slow secular change of zero in the mercury thermometer is then of little moment , while great importance attaches to the accuracy of the individual readings .
Failure of the Box Mercury Thermometer to give the True Temperature of the Coils and Bridge Wire .
S 25 .
If the room temperature is below 20 ' C. , and artificial heating is employed , we may have a practically stationary reading on the box thermometer , and yet find it differing by several tenths of a degree from a second thermometer whose bulb is at a different level in the coil chamber .
Under such circumstances the temperature inside the coil chamber also varies with the distance from the side or end walls .
The coils are of various shapes and diameters , some coming much nearer the top and bottom of the coil chamber than others , and they are necessarily at different distances from the ends .
Thus , under the conditions specified above , different coils may possess different mean temperatures , and the temperature given by the mercury thermometer may differ sensibly from that of any one of the coils .
Even with no artificial heating of the box , the temperatures at different parts of the coil chamber may differ by several tenths of a degree , when the room temperature is changing moderately fast .
In such a case the presence of water in the protecting tank seems a positive drawback , as it makes the temperature of the coil chamber lag behind that of the rest of the room .
When the temperature of the coil chamber is altering , whether through artificial heating or otherwise , the various coils doubtless alter 28 Dr. C. Chree .
Investigations on their temperature at different rates , and unless the mercury thermometer is specially chosen it may have a greater or a smaller lag than any one of the coils .
The mercury thermometer used in the Kew box , like most open-range thermometers , contains a considerable quantity of mercury , and when the temperature changes fast it lags very sensibly behind the coils .
For instance , in some special steam point observations with Ki , when the recorded box temperature rose 2 ' *5 in twenty minutes , the corrected bridge-wire reading fell about 0'*06 , instead of remaining practically constant as it ought to have done .
The figures really show that the temperature of the box thermometer had lagged about 0'*7 C. behind that of the coils in use .
This of course was an extreme case , and under ordinary experimental conditions the difference between the temperatures of the box thermometer and the coils is hardly likely to exceed 0''3 , and is probably seldom half this .
Necessarily , however , a good deal depends on the depth of the bulb of the thermometer inside the coil chamber , and any alteration of that depth should be carefully avoided .
S 26 .
The temperature of the bridge wire is open to much greater uncertainty owing to its relatively exposed position .
The whole bridge wire resistance in the Kew box is fully 30 box units , and if this were all on the platinum thermometer side of the Wheatstone bridge the error arising from the assumption that it possesses the temperature of the box thermometer would be approximately 0-008r , where r is the error , in degrees Centigrade , in the temperature assigned .
As r may easily amount to several degrees , an error of several hundredths of a degree might easily creep in from this cause if one used a large fraction of the bridge wire .
Theoretically it is possible to keep the bridge-wire reading within 5 units of the bridge centre , * thus reducing the uncertainty to less than one-sixth of that existing in the extreme case supposed .
This , however , has not been compatible with our practice of employing the same coil combinations for all the thermometers of the same pattern .
Even with the drawback of temperature uncertainty it is often advantageous to make a liberal use of the bridge wire .
It facilitates finding the balance , and in work such as the comparison of mercury thermometers at high temperatures , where rapid reading is essential , it is a great convenience .
It is thus highly desirable that the bridge wire should be better protected than in the Kew box , and that its temperature should be measured directly .
Insufficient Immersion .
S 27 .
Supposing a correct mercury thermometer to be immersed in ice up to the reading \#151 ; 10 ' C. on the stem , there is a column * This was the policy recommended by Mr. Griffiths originally .
Platinum Thermometry at Kew Observatory .
29 of mercury of about 10-degree divisions exposed to the influence of the surrounding atmosphere , whose temperature is , say , 20 ' C. This emergent column possesses a mean temperature somewhere between 0 ' and 20 ' , and so occupies a larger volume than if all were at 0 ' ; on this account alone the thermometer must read too high .
The emergent column has a second influence tending in the same direction ; it serves , conjointly with the glass tube , as a path for the conduction of heat to the bulb .
In high-temperature measurements the error due to a long emergent column may amount to ' several degrees , the thermometer in this case reading too low .
The conduction along the stem is usually much the less important influence of the two .
A platinum thermometer is not wholly free from immersion difficulties .
It is not sufficient to have the spiral inside the bath or chamber whose temperature is desired , so that the defect in a platinum thermometer is most analogous to the second source of error described in the mercury thermometer .
If , for instance , a platinum thermometer is buried in ice to only a little over the top of the spiral , the tube emerging in a room at 20 ' C. , the heat communicated to the spiral through the surrounding air or through the connecting wires raises the reading appreciably .
Similar insufficient immersion in the hypsometer sensibly lowers the reading .
It was difficult to settle on a suitable basis-for comparing the effects of insufficient immersion in the different platinum thermometers .
The spiral is wound on a mica frame , and usually one of the small mica discs which hold the wires apart is situated immediately above the top of the frame .
When this occurred I have taken the " bulb " of the thermometer as extending to the first mica disc ; but in one case , where the interval between the first disc and the frame was considerable , I took instead the top of the frame itself .
As a rule , the spiral stops short of the top of the mica frame by about f cm .
; so the definition is at least a lenient one to platinum thermometers .
The tubes of K5 , K6 , and K7 being of transparent glass , the length of the " bulb " was visible from outside ; but in Ki and K4 the tubes had to be taken off to allow of the requisite measurements being made .
In the immersion experiments in steam the thermometers were carried in the usual way by a cork , about 3'25 cm .
long , fitting the neck of the hypsometer , and the immersion was counted up to the lower face of the cork .
Thus fully 3 cm .
of the tube immediately above the portion exposed to the steam was protected from direct cooling by the air .
S 28 .
Table X gives particulars as to the results obtained .
It includes the errors observed with the various immersions , measured on the hypothesis that an immersion of " bulb " + 10 cm .
was in all cases sufficient .
It also gives the lengths of the " bulbs , " and the mean temperature in the room during each series of experiments with each oo o Table X.\#151 ; Errors due to Insufficient Immersion .
Thermo- Temperature of air .
Length of " bulb .
" Ice experiments immersion .
Steam experiments immersion .
meter .
Steam expts .
Ice expts .
Bulb + 0 .
+ 1J cm .
+ 2\ cm .
+ 5 cm .
-f 7i cm .
Bulb + 0 .
+14 cm .
+ 2| cm .
4- 5 cm .
+ 7\ cm .
o 11 o 18 cm .
5 6-183 6*068 0 033 6-014 6-oo7 6-216 0 .
085 6-025 o'-ooo o-ooo 15 15 4f 0-555 0-223 0-095 0'012 0-007 0-904 0-423 0-191 0-049 o-ooi 16 18 4 0-385 0-224 0-110 0-026 0-004 0-547 0-248 0-103 0-022 o-ooo K6 13 15 6* 0-154 0-105 0-071 0-022 0-009 0-320 0*186 0 095 0-029 0-007 K ; 14 16 4* 0-284 0-167 0-093 0-024 0-007 0-448 0-251 0-134 0 047 0-015 Means ... 14 16 5 0-313 0-157 0-080 0-020 0-007 0-487 0-239 0-110 0-029 0-005 Ratios to error when " bulb immersed t T t " only j 1 : 0-50 : 0-26 : 0-06 : 0-02 1 : 0-49 : 0-23 : 0-06 : 0 01 Dr. C. Chree .
Investigations on Platinum Thermometry at Kew Observatory .
31 thermometer .
This last information is of most consequence in the case of the experiments in ice , because the errors must then increase as the temperature of the room rises , and during the immersion experiments the room temperature was much below its ordinary summer value .
Some of the figures are based on two or three complete independent experiments , but others depend on only one experiment .
In most cases there were in addition several check experiments , with only two or three different immersions , but I have left these out of consideration in constructing the table .
The results cannot claim any very high degree of precision , because it is difficult to secure uniformity in the conditions , more especially when the immersion is in ice .
The difficulty is greater the less the immersion .
When the " bulb " only is immersed in ice , the reading after attaining a minimum usually rises somewhat .
The phenomenon seems mainly due to melting of the surface ice , with consequent deterioration of contact between it and the thermometer ; but possibly a difference in the circumstances under which the thick platinum leads and the compensator wires cool may possess some influence .
In taking the minimum readings , as I have normally done , I have of course , if anything , favoured the thermometers .
This has also been done to some small extent in assuming that " bulb " + 10 cm .
is in all cases ample immersion .
As a matter of fact , an unmistakable difference was observed in the reading of several of the thermometers when the immersion was increased from " bulb " + 10 cm .
to " bulb " + 12-5 cm .
This difference was not , however , always in one direction , the phenomena in one case distinctly suggesting that cooling an additional 2'5 cm .
of the compensator wires had more effect than cooling an additional 2-5 cm .
of the platinum leads .
The experiments are not sufficiently varied to justify any too confident explanation of the causes of the large differences between the different thermometers .
Thus the pre-eminence of the errors in K4 might at first sight be attributed to the fact that the tube is of porcelain and of exceptionally large diameter ; but Ki is also of porcelain , while Kg has a much larger diameter than Ky. Perhaps the most probable explanation is that K4 is short for its tube , so that the mica frame\#151 ; which is exceptionally short\#151 ; only comes to within about 1 cm .
of the end of the tube .
Ceteris paribus , we should expect that with a given immersion , the error would vary approximately as the difference between the temperature of the immersed portion of the tube and that of the air of the room , but to confirm or refute this conclusion would require a considerable number of careful experiments in which either the temperature of the bath or that of the room alone is varied , all the other conditions remaining unchanged .
We cannot make use of the experiments in ice and steam for this purpose , because the conditions under which the transfer of heat occurs are radically different .
We can meantime only 32 Dr. C. Chree .
Investigations on note as an isolated fact that the immersion errors were only about 50 per cent , greater in steam than in ice , whereas the difference between the temperatures of the room and the immersed portion of the thermometer was about six times as large in the former case as in the latter .
It may be only a coincidence\#151 ; but if so it is a very curious one\#151 ; that in the mean results , both in the ice and steam experiments , the errors diminish in geometrical progression as the immersions increase in arithmetical progression .
An addition of D25 cm .
( ie .
, one-quarter the length of the " bulb " ) lowers the error almost exactly one-half .
S 29 .
Returning to the general question , we see from Table X that the immersion should be at least 10 cm .
, in addition to the length of the " bulb " ; in all , in the Kew thermometers , at least 15 cm .
( 6 inches ) .
If the whole length immersed is less than 6 cm .
, the error in an ice or steam point experiment is unlikely to be much under 0'T C. ; and until the total immersion exceeds 10 cm .
, the error is likely to exceed 0'-01 C. In " fixed point " experiments it is usually easy to have an immersion exceeding 15 cm .
, and the main results dealt with in this paper are , I believe , free from appreciable uncertainty on this ground .
There are , however , other circumstances under which immersion difficulties arise , as for instance in the comparison of mercury and platinum thermometers at high temperatures .
Unless both the material of the bath and the contained liquid are transparent , we must have the divisions of the mercury thermometer which we have to read emergent .
But with ordinary stirring it is also essential , except in a specially protected bath , that the bulb of the mercury thermometer and the spiral of the platinum thermometer should be about the same level .
These conditions are often inconsistent with sufficient immersion of the platinum thermometer when we are comparing the lower part of the scale of the mercury thermometer .
In fact , my attention was first directed to the question of immersion of platinum thermometers through the unexpected result that raising equally the height of Kr and a certain mercury thermometer in a bath of molten metal lowered the reading of K7 most .
Another case where insufficient immersion is to be feared is in the determination of melting points of metals such as silver .
The crucible supplied for use with melting silver at Kew would , if full to the lip , allow an immersion of only 10 cm .
in all As the silver is stirred during part of the experiment , and is too precious a material to splash all over the furnace , the immersion cannot well exceed 9 cm .
, and may be considerably less .
It is by no means impossible that insufficient immersion may have Platinum Thermometry at Kew Observatory .
33 been accountable for ' the fact that the mean values of the melting point of silver from experiments by Mr. Hugo and myself ( six in all ) were only 958''2 C. with Ki and 9570-5 with K2 ; while Mr. Heycock and Mr. Neville , at Cambridge , found 961 '*1 C. with the same sample of silver but a different thermometer .
This is 2 ' above the highest individual observation at Kew Observatory .
Slowness of Platinum Thermometers in the True Temperature .
S 30 .
When a mercury thermometer is taken from a room at one constant temperature into another room at a different constant temperature , some time elapses before the reading becomes steady .
while , if the temperature of the surrounding medium alters , the thermometer shows a lag which varies with the thickness of the glass walls and the amount of mercury in the bulb .
As we have seen , the naked coils in the Kew resistance box have a smaller lag than an ordinary mercury thermometer , but it is otherwise with platinum thermometers .
Here the wire does not come into direct contact with the external medium , and the attainment of the steady state is somewhat slow .
Experiments have been made with some of the Kew thermometers which have been suddenly transferred from a bath about 15 ' C. into ice , or into a hypsometer in which the water is freely boiling .
The interval , in seconds , was observed which elapsed before the bridge-wire reading came to 1 ' , 0''5 , or other fixed distance of the final stationary position ; this position had been found in a preliminary experiment .
In all cases the immersion was similar to that of the ordinary fixed point experiments .
Two mercury thermometers , Nos. 686 and 750 , were experimented on in an exactly similar way for the sake of comparison .
They are both Kew standards of the following dimensions :\#151 ; Length of fundamental interval .
Distance of zero mark from lower end of bulb .
Length of bulb .
Diameter of bulb .
cm .
cm .
cm .
cm .
K.S. 686 ... .
35 21 2 -5 0*6 K.S. 750 1 32 8 2-2 0 *5 I have not thought it necessary to go into full details of the observations .
The following table gives the number of seconds required to get to 1 ' and 0'-5 of the stationary temperature in ice , and to 1 ' , 0o,5 , and 0o,25 of the stationary temperature in steam .
It was found that VOL. LXVII .
D Dr. C. Chree .
Investigations on the rapidity of attaining the steady state in ice depended greatly on the tightness of the packing and the moistness .
Thus , particulars are given of experiments made with the ice variously treated\#151 ; Table XI .
Thermometer ... Ice .
Steam .
As usual , but not packed tight .
As usual , packed tight .
Freshly moistened .
Mixed with a good deal of water .
1 ' .
0'-5 .
1 ' .
0'-5 .
1 ' .
0 ' *5 .
1 ' .
0'*5 .
1 ' .
0'-5 .
0'-25 .
K , 51 38 50 33 44 28 35 37 44 58 Kt 103 132 50 58 44 53 \#151 ; \#151 ; 51 61 k6 \#151 ; \#151 ; 48 61 \#151 ; \#151 ; \#151 ; \#151 ; 65 77 89 k7 \#151 ; \#151 ; 31 44 \#151 ; \#151 ; \#151 ; 1 43 \#151 ; 64 K.S. 686 \#151 ; \#151 ; 9 13 \#151 ; \#151 ; \#151 ; \#151 ; 1 9 K.S. 750 \#151 ; 11 20 \#151 ; \#151 ; \#151 ; \#151 ; 6*5 \#151 ; \#151 ; j Most of the data are based on at least two experiments .
The different experiments with any one thermometer in steam were in good agreement .
The agreement in the ice experiments was pretty fair when the ice was moistened or was tightly packed ; when , however the ice was not freshly moistened and was only loosely packed the results were very variable .
The intervals 1 ' , 0o,5 were generally only approximate , and to elucidate the exact law of cooling or heating would require a more complete investigation .
The data are , however , exact enough for general conclusions .
The most striking fact is the extreme slowness of all the platinum thermometers as compared to the two mercury thermometers ; and it should be noted that these two thermometers are not of a type intended to be rapid .
To get within 1 ' of the steady reading , the temperatures of the thermometers had to alter about six times as much in steam as in ice .
The times required for these two changes are pretty much alike in all the platinum thermometers , supposing the ice tightly packed , as it usudly is in the normal experiments .
Thus the approach to the steady state is decidedly slower in ice than in steam , and a longer time must be allowed in the former case than in the latter .
S 31 .
Some rough calculations , based on the observations with Ki , summarised in Table XI , point to the conclusion that 3'5 minutes ' immersion in ice and 2 '5minutes ' immersion in steam would suffice to bring the reading within 0''0005 of its stationary value in the case of this thermometer .
K4 and K6 are decidedly less rapid , and I should conclude that ten minutes ' immersion in well-packed ice and five minutes ' exposure to steam are by no means very excessive allowances in the Platinum Thermometry at Kew Observatory .
35 .case of an average platinum thermometer .
It has been customary at Kew to allow a fully longer immersion than ten minutes in ice before taking the first reading , and I think that in the case neither of the ice nor the steam point observations has there been any sensible error through excessive hurry .
At the sulphur point the attainment of the steady state is a tedious process .
Readings taken with the short-stem thermometers K3 and K4 within thirty or forty minutes of their first exposure to sulphur vapour had nearly always to be rejected , being conspicuously low .
One had in fact to allow about an hour with these thermometers .
Even with the long-stem thermometers , whether glass or porcelain , at least forty minutes ' exposure to sulphur vapour was found desirable .
A good deal depends on the gas supply and on the quality of the Bunsen burner used .
In some of the sulphur point experiments at Kew it is open to doubt whether the stationary temperature had been absolutely attained , but I do not think any serious uncertainty was introduced in this way .
Impurity in the Ice used in getting R0 , or Variability in the Method Treating it .
S32 .
The ice employed at Kew Observatory is supplied in large blocks , which are washed prior to use .
As the testing of thermometers in ice is a frequent occurrence , the assistants have no lack of experience , and since a planing machine was introduced some years ago the ice as prepared has been very uniform and finely divided .
The ice employed in the platinum thermometry has been taken from the supply used in the ordinary test work , and as the purity of this is frequently checked by observations with Kew standard thermometers no serious impurity need be feared .
The check would , I allow , be hardly adequate to settle a question of two or three thousandths of a degree , and no doubt for work of the highest accuracy a more stringent test would be desirable .
Strong confirmatory evidence of the uniform purity of the ice is afforded , as will be seen presently , by the small variability in the observed values of the fundamental intervals of the thermometers .
The method of preparing and moistening the ice seems to exert a small but sensible influence on the depressed zero readings of mercury thermometers , possibly because the amount of the depression must be influenced by the rate of cooling .
There seems , however , to be very little if any effect on the zero reading of a platinum thermometer , so long as the immersion is ample and the temperature of the room is moderate .
Error in the Barometer .
S 33 .
Ordinary errors of reading , lag , or improper temperature .correction would merely introduce irregularities into the calculated Dr. C. Chree .
Investigations on boiling points ; but an undetected change of zero in the barometer , or an alteration in the method of reduction , would simulate a change in Ki and R " the same for all the thermometers .
After the first few observations the Royal Society 's old double tube mercury barometer\#151 ; repaired some years previously by Messrs. Negretti and Zambra\#151 ; was set up in the platinum thermometer room , and it has since then been used in all the steam and sulphur point observations .
A recent comparison between it and the Observatory standard barometers shows that no certain change of zero has occurred since it was set up .
The surface of the mercury in the cistern of the barometer is about 6 inches lower than the surface of the water in the hypsometer , and about 11 inches lower than the surface of the boiling sulphur .
No allowance has been made for these differences of level , as they would introduce extremely small and practically constant errors in the two boiling points .
The same reduction tables have been in use throughout , and the constant factor 1-0006 has been applied in the reduction to gravity at latitude 45 ' .
Until standard gravity is more exactly defined , and relative values of gravity more carefully determined , the accuracy of the reduction is open to question , but for the purposes of the present inquiry the uncertainty is quite immaterial .
As to the accuracy of the individual barometer observations , I need only say that readings by Mr. Hugo and myself seldom differed by more than 0-001 inch .
The instrument has a tube of 0-6 inch in diameter and responds rapidly to alterations of pressure , so that no serious error need be apprehended from lag .
Impurity in the Sulphur , \amp ; c. S 34 .
The sulphur made use of in the experiments has all been obtained from Messrs. Baird \amp ; Tatlock ; it has been produced by Chance 's process .
Though several supplies have been obtained , at different times , no discontinuity has been observed in the sulphur points ; thus , unless our experience has been unduly favourable , sulphur has at least one great merit as a medium for supplying a fixed high temperature .
There are , however , certain disadvantages which have introduced some variability into the conditions of experiment .
After an experiment the sulphur solidifies in the tube containing it , and the remelting it on a subsequent occasion not infrequently ends in cracking the tube .
On one or two occasions a crack went right round the tube , at a considerable height above the bulb , after the sulphur had been melted and experiments were progressing , and on other occasions the tube was found to have cracked a day or two after the sulphur had cooled .
Of one batch of tubes , of thicker glass than usual , several flew into pieces after the sulphur had liquefied , and the observer was luck } Platinum Thermometry at Observatory .
37 in escaping with no damage except to his clothes .
In consequence of this defect , we have had to employ about a score of tubes , and these have not been strictly uniform in length , thickness , or diameter .
We have also had to employ a considerable number of small asbestos cones , of the pattern recommended by Mr. Hey cock and Mr. Neville , and the tightness with which these have fitted the tubes has varied sensibly .
Thus the sulphur point observations have been taken under more variable conditions than those at the steam point .
We have not indeed observed any certain consequences to follow any of the minor changes described above , but more careful experiments would be necessary to justify the conclusion that the variations are immaterial .
I am , in fact , inclined to think that variations in the length of the asbestos cones , or in the tightness with which they fit the tube , are sufficiently probable sources of uncertainty to deserve investigation .
Error in the Formula assumed for the Variation of the Boiling Point Sulphur with Pressure .
S 35 .
In another direction there is , I think , little doubt that appreciable error exists , viz. , in the reduction of sulphur point observations to the standard barometric pressure .
Professor Callendar and Mr. Griffiths accept a simple linear relation M = ( p - 760 ) x 0082 ... ... ... ... ... ... . .
( 10 ) , where M is the increment in the temperature of sulphur vapour , on the air scale , when the barometer is p mm. instead of 760 .
This is based on some experiments by Regnault , who examined a wide range of pressure , but had so few experimental points that interpolation is very uncertain .
A priori , having regard to the corresponding phenomenon in steam , one would hardly anticipate great accuracy from a linear relation , unless restricted to a very limited range ; while in ordinary every-day use we must be prepared to cover at least a range of 30 mm. The Kew experiments , having been taken without any special regard to the atmospheric pressure , naturally cover a fairly wide range , and thus afford a favourable opportunity of testing the accuracy of ( 10 ) .
To this end I have carefully reduced all the sulphur point experiments in a uniform way .
Assuming Callendar 's formula t-pt = 8[(t/ 100f-(t/ 100 ) ] , we may , near the sulphur point , conveniently give it the form pt-pts = ( t- ts ) [ 1 - 100-2S(2* !
s - 100 ) ] - 100~28(\#163 ; - ts)-. .
( 11 ) , Dr. C. Chree .
Investigations on where ts and pt , are the temperatures , on the gas and platinum scales , of the sulphur boiling point at standard pressure .
Hence if ( 10 ) be correct , we have , pt answering to the temperature of sulphur vapour when the pressure is p mm. , pt8 = pt -0-082 ( p-760 ) [ 1 -100-28(2^-100 ) - 100~2S x 0-082 -760 ) ] ... ... ( 12). .
Supposing , for instance , 8 = 1-5 , and = 444 ' '53 , then pts = pt-0-082( p- 760 ) [ 0-88164 - 0-000012 - 760 ) ] .
For any ordinary barometric range the term containing^ - 760 inside the square bracket is negligible , and so we may take .pt , =^-0*0723(^-760 ) ... ... ... ... ... ... .
( 13 ) , where the last significant figure is , of course , uncertain .
In actual application pt is the observed value of 100(E - E0)/ ( K1 - E0 ) in a sulphur point experiment taken when the reduced barometer reading is p , while pt , is the value calculated from this experiment for the platinum temperature of the boiling point of sulphur at standard pressure , on the assumption that ( 10 ) is correct .
S 36 .
In calculating pt , I adopted two methods : First I took for Ri - Eo in each group of experiments with a given thermometer ( a including all the experiments made during one year , or during the time one particular calibration was in vogue ) the mean of the observed values of the fundamental interval during the epoch in question .
Next I used for each sulphur experiment the individual value of Ei - E0 found from the corresponding ice and steam point observations .
These observations were , usually taken on the same day as the sulphur point observation , or on the previous day .
I thus obtained two series of values of pt , .
In the first , one principal source of uncertainty was .
largely eliminated ; while , in the second , another principal source of uncertainty was largely reduced .
The following table shows the results for the longest epoch considered , viz. , from September , 1897 , to January , 1899 , inclusive , covering the time when Calibration III was in vogue .
During a set of sulphur experiments with the different thermometers the barometric height of course varied somewhat , thus the pressures given in the first column of the table are only approximate .
The break in the table separates the cases where the reduced barometric height was above 760 mm. from those in which it was below .
The results in the last two lines are based on the exact barometric readings observed during the experiments on the individual thermometers :\#151 ; Table XII .
Approximate barometric height in mm. Values of pts using a mean fundamental interval for the group .
Tallies of pt$ using individual values for the fundamental interval .
K , .
k3 .
k4 .
k7 .
Kn k3 .
k4 .
k7 .
421 ' + 421 ' + 421 ' + 421 ' + 421 ' + 421 ' + 421 ' + 421 ' + 776 .61 .81 .80 .65 .59 .76 .74 .56 773 .61 .63 .69 .58 .60 .63 *71 .56 769 .43 .60 .58 .52 .52 .63 .50 .57 767 .68 .56 .62 .51 .55 .52 .61 '48 766 .63 .72 .69 .65 .62 .69 .64 .44 765 .49 .61 .60 .53 .56 .63 .69 .59 763 .41 .48 .49 \#151 ; .46 .47 .53 - .
758 .50 .48 .47 .42 .47 .57 .55 .38 756 .45 .44 .46 \#151 ; .53 .44 .47 \#151 ; 753 .37 .52 .42 .41 .43 .52 .41 .37 751 .44 .57 .64 \#151 ; .41 .57 .60 \#151 ; 746 .48 .55 .61 *46 .43 .58 .59 .46 Mean when pressure \gt ; 760 , .551 .630 .639 .573 .557 .619 *631 .533 \gt ; \gt ; \#187 ; \gt ; \gt ; ^ 7b0 .
.448 .512 .520 .430 .454 .536 .524 .403 Differences .
.103 .118 .119 .143 .103 .083 .107 .130 Difference per 1 mm , . .
" between .0064 769 -5 753 -5 .0079 768 -2 753 2 0079 767-8 752 -8 .0080 768*8 mm. 750-9 " .0064 .0055 .0071 .0073 S ' \#163 ; SL 0 !
CO CO Dr. C. Chree .
Investigations on S 37 .
A similar treatment of the entire number of observations with these four thermometers\#151 ; viz. , twenty-four with K3 , twelve with K7 , and twenty-one with both Ki and K4\#151 ; making use of the individual fundamental intervals only , gave the following results :\#151 ; Table XIII .
Pressure .
Mean pressures 700 + Mean values of pts , 421 + K , .
K3 .
K4 .
K- .
Ki .
K3 .
K\lt ; .
k7 .
Above 760 ... .
67 -57 66 -66 66 -19 68 *77 .508 .612 .625 '533 Below 54-08 54-06 54-63 53-91 .468 .485 .526 .402 Differences ... .
13-49 12-60 1156 14-86 .040 .127 .099 .131 Thus variation in pts per 1 mm. of pressure 0 -0030 0 -0101 0 -0086 0-0088 As we shall see later , the thermometer K4 has undergone a change .
Its pts has risen , and the apparent smallness in the variation in its case in Table XIII seems largely due to the fact that amongst the latest experiments there was a preponderance of cases where the barometer was below 760 mm. S 38 .
Tables XII and XIII prove , I think , beyond a doubt that the Callendar-Griffiths reduction formula ( 10 ) assigns an appreciably too small value to the variation of the temperature of sulphur vapour with pressure near 760 mm. This is the conclusion to which the results from every single thermometer point , and the phenomena are too conspicuous to be attributed to accident or to experimental errors .
The agreement between the results in Tables XII and XIII is , I admit , not all that might be desired , and any numerical conclusions based on them are perhaps hardly likely to be final .
Still I have thought it worth while to see what these conclusions are , proceeding as follows : \#151 ; Suppose that the Callendar-Griffiths formula ( 10 ) is defective in omitting a term in ( p \#151 ; 760)2 and in assigning an incorrect value to the constant multiplier of the term in p - 760 .
Then , supposing as before 6 = 1 *5 , we should replace ( 13 ) by pt = pts+ ( 0-0723 +x ) ( p -760 ) +y(P~ 760)2 ... . .
( 14 ) , where x and y must be determined from the experiments .
As all the observations had been already worked up applying ( 10 ) , it seemed simplest to take for data the values calculated in this way for pts , determining x and y by least squares from the observed differences between the calculated values and their mean .
Platinum Thermometry at Kew Observatory .
41 I first took the data in Table XII and added to the individual values for pts in Ki , Iv3 , and Kr the mean of the differences between the values of ptsfound with these respective thermometers and the corresponding values in K4 .
As the corresponding values of pts in these four thermometers answered to nearly the same barometric pressure , one got in this Avay a mean pts answering very fairly to the mean barometric pressure .
Applying formula ( 14 ) to the figures so obtained , I obtained by least squares x = +0-0063 , y= +0-00018 ... ... ... ... ... . .
( 15 ) .
Next I took all the observations made with K3 and K4\#151 ; excluding one when the asbestos cone fell off\#151 ; and combining the observations made on the same days got twenty-four values of pts as given by the formula ( 10 ) .
Combining these two and two in the order of the corresponding barometric pressures , I found from the resulting twelve data by least squares $ = +0-0087 , y= +0-00021 ... ... ... ... ... ... ( 16 ) .
For ready comparison I show side by side the formulae resulting from ( 15 ) and ( 16 ) , as well as those based on the Callendar-Griffiths value .
In these formulae , ts and pt$ are , as before , the temperatures on the air and platinum scales of sulphur vapour boiling under the standard pressure 760 mm. , while t and pt are the corresponding temperatures when the pressure is not 760 , but p mm. It is supposed that 1*5 in the formulae for pt .
Callendar and Griffiths , { pt = pts + 0072 - 760 ) , from Regnault .
1 I = 4 + 0"082(^\#187 ; -760 ) .
From ( 15 ) , i.e. , by experiments on Ki , K3 , K4 , and K^ , between Sept. 1897 and Jan. | pt = pts+0-078(p - 760 ) + 0-00018 - 760)2 , \gt ; t = ts + 0-088 ( p-760 ) + 0-00020 ( p-760)2 .
1899 .
From ( 16 ) , i.e. , by ex-"A periments on K3 and ly \#151 ; K4 between Oct. 1895 and Sept. 1899 .
pt8 + 0-081 ( p - 760 ) + 0-00021 ( p - 760)2 , ts + 0-092 ( p- 760 ) + 0-00024 - 760)2 .
In restricting the second calculation to K3 and K4 I was guided by the obvious change in K4 and by the comparative fewness of the experiments at the sulphur point with K-_\gt ; , K5 , and K- .
According , however , to Table XIII the variation of pts , per 1 mm. pressure , deduced from the experiments on K3 , is exceptionally high .
Further , the agreement between observed and calculated values in the second calculation was not so good as in the first .
I am thus disposed to attach somewhat Dr. C. Chree .
Investigations on more weight to the result of the first calculation .
In any case I have little doubt that the formula t \#151 ; t8 + 0-090(^-760 ) + 0-0002 ( ^-760)2 ... . .
( 17 ) will prove considerably more exact than the formula ( 10 ) now in use .
Further direct experiments on the point are , in my opinion , desirable .
S 39 .
In our previous calculations we have employed Callendar and Griffiths ' value , 444 ' *53 O. , for the boiling point of sulphur .
This is supposed to be measured on the scale of the constant pressure air thermometer .
It is about 0o\lt ; 7 C. lower than the value recently found by Chappuis and Harker , employing the constant volume nitrogen thermometer .
We do not as yet know enough about gas thermometry to judge of the consistency of the two results , or of the degree of accuracy either can claim .
It is thus desirable to know how the previous results would be influenced by a slight change in the accepted value of t8.Such a change would not affect any formula like ( 10 ) connecting t and p , and it would modify formulae like ( 13 ) containing pt only indirectly through change in S. The change AS in S , answering to a small change Ats in ts , is given by AS = A ts[104 - 8(2 ts- 100 ) ] - - 100 ) ] .
Using Callendar and Griffiths ' value of tg we get approximately , when 8 = U5 , AS = 0-0575A4 .
For instance , if A ts \#151 ; 07 ' C. , then AS = 004 , approximately .
Constancy or Variability of Platinum Thermometers .
S 40 .
Supposing the R0 of a thermometer , determined in the usual way , to show a distinctly lowered value as time proceeds , the cause might be a change in the leads , or in the proportional arms , or in the box coils , and not a real change in the thermometer itself .
There are no certain data as to the original equality of the proportional arms ; but as Dr. Harker recently found them to differ by only about three parts in 20,000 , and as they were , doubtless , made as nearly as possible equal at first , there is reason to suppose that any uncertainty on this head is small .
On the other hand , as shown by Table III , changes in the leads have been large and fluctuating .
As no data exist for eliminating this uncertainty in a direct way in the earlier observations , I have based the inquiry into the constancy of the thermometers mainly on an intercomparison of themselves .
As the same coil combinations were used for every one except Ko , and as the thermometers have nearly equal Platinum Thermometry at Kew Observatory .
43 fundamental intervals , and were compared in the great majority of instances on the same or on successive days , the differential results are hut little exposed to the chief sources of uncertainty .
Before discussing the differential results I shall glance briefly at the absolute data obtained when the leads were interchanged .
A summary of these is given in Table XIY , along with results of observations without interchanged leads on October 4 and December 8 and 9 , 1897 :\#151 ; Table XIY.\#151 ; Values of R0 .
Date .
Leads .
Ki .
k3 .
k4 .
Kr .
September 21/ 1897. . . .
interchanged 257-756 257 -856 257 -453 256 -708 October 4 , 1897 as usual .739 .833 .431 .672 December 8 and 9 , 1897 as usual .750 .827 .420 .672 Mean of experiments between May and September , 1899 interchanged .723 .856 .462 \#166 ; m The leads had been adjusted to approximate equality between September 21 and October 4 , 1897 ; we may thus reasonably suppose that the changes experienced prior to December 9 , 1897 , were small .
Also , judging by what was observed on October , 1898 , such change as might have occurred would tend to diminish R0 .
We may thus regard the results obtained on October 4 , and December 8 and 9 , 1897 , as proving that the observations on September 21 did not suffer from any notable source of error peculiar to the day .
Comparing the results obtained on September 21 , 1897 , with the mean results for 1899 , we obtain a surprisingly close agreement in the case of every thermometer except Kj .
This certainly is pretty strong evidence of constancy in at least three of the thermometers during the last two years .
It is of course possible that a common change might have occurred , and that it was masked by a corresponding equal change in the box coils ; but as these coils are of a different material , platinum silver , and have been exposed to wholly different conditions , such a contingency is improbable .
S 41 .
For the purpose of intercomparison , I have selected as standards of reference K3 and K4 .
These two thermometers are almost identical in pattern , and have been exposed to almost exactly the same treatment , usually being compared in immediate succession to one another .
Taking the mean R0 , Ri , and Rs given by these thermometers on each occasion , I have found the difference between these means and the R\lt ; \gt ; , Ri , R* of each of the other thermometers .
I preferred taking the mean for the two thermometers to taking either alone as a standard , because Dr. C. Chree .
Investigations on by doing so , observational errors should be rendered less important .
Also there was no reason for preferring one to the other .
The R0 and Ri of K(j have been multiplied by 04 to reduce them to approximate equality with the corresponding quantities in the other thermometers ; and in this instance the data given are limited to the cases where the leads were interchanged .
The reason for this is that the value of R0 in K0 is affected by a given inequality in the leads by no more than the Ro in K3 or K4 , so that inequality in the leads would influence such a quantity as 04 R0 ( in Ko ) - ( mean R0 of K3 and K4 ) .
The differences between the Rq , Ri , and Rs of K3 and K4 are also given , so that the differences of any thermometer from K3 and K4 separately .could be at once obtained from the table .
Such a heading as K3 - K4 means that the Rq , Rj , or Rs.\#151 ; as the case may be\#151 ; of K4 , when subtracted from that of K3 gives the results in the columns of Table XV \lt ; p. 45 ) .
The data enclosed in the [ ] brackets are , I have reason to think , affected by experimental error , while those in the ( ) brackets are , I believe , correct but due to exceptional circumstances .
These data are not taken into account in the mean values .
Data in the brackets { } refer to two independent observations during the same month .
The blanks in the case of K2 and K5 arise from the absence of these thermometers at Sevres and from the changes in the former thermometer during 1896 .
S 42 .
Table XV shows that if any relative change has occurred between K3 and K4 since 1895 it must be very small .
The figures are not , however , absolutely inconsistent with a slight reduction in the difference between the two resistances .
There is the clearest evidence of a gradual fall in the Ro and Ri of Ki relative to the corresponding quantities in K3 and K4 , the change being greatest in the case of Ri .
On the other hand , the Rs of Ki has apparently risen during the last two years relative to the mean Rs of K3 and K4 .
As the resistance of K4 at 0 ' C. is intermediate between the resistances of K3 and K4 , it is impossible to attribute the phenomena to changes in the proportional arms , or in the coils , or to any other external cause .
I thus see no alternative to the conclusion that in Ki there has been a diminution in the resistance in ice and in the fundamental interval , accompanied by an increase in the ice-sulphur and steam-sulphur intervals .
In its early days K4 was several times exposed to the temperature of molten silver , but the last exposure preceded 1897 .
Table XV shows no other certain example of relative change except in the case of K5 when quite new .
The first experiment made with this thermometer was on February 12 , 1896 , by which time the observers had become fairly Table XV.__Differences between Resistances of Thermometers .
Date .
k3-k4 .
Ki- ^(K3 + K4 ) .
K3- |(K3 + K4 ) .
K5- i(K3+K4 ) .
0*4K6\#151 ; ^(Kjj + K^ ) .
J(K3+K4)- k7 .
1895 .
lb- 0'415 Ri- 0-675 K* .
K0 .
Ri- B* .
Bo* R " Ro- Rl B* .
B0 .
Ri- B0 .
Ki .
Its .
0-493 0*735 1*526 1896 .
0*437 0*633 [ 1*860 ] 1*514 1*535 \#151 ; \#151 ; \#151 ; - - ( 0-786 ) ( 1-154 ) n-A8S 1-082 2*482 2*396 March\#151 ; April 0*419 0*404 0*696 0*672 0*250 0218 0*148 0*172 0*012 \#151 ; \#151 ; - 0*658 0*643 0*632 0*629 1*054 1*041 1*030 2*319 2*291 Jiinp 0*398 0*667 1*520 0*216 0*203 '*'* 2*351 * August October 0-393 0-378 0-656 0-609 1*553 0*193 0*180 0-168 0-157 - - - 1*020 *\#151 ; \#151 ; 0*971 1-526 Mean for 1896 ... 0*405 0*665 1*530 0*210 0*186 0*111 - - - 0-649 1*045 2*368 \#151 ; 1897 .
March 0-397 0-401 0-673 0-659 - - 0-128 0-137 0-122 0-118 - - - - 0*637 1*028 - - - 0*945 0*950 0*972 1-526 1-542 [ 1-621 ] 1-465 3-239 0-448 0*621 1*408 0*193 0*046 0*019 \#151 ; \#151 ; 0-436 0-712 0*946 3-308 0*403 0*673 1*565 0*102 0-037 0*053 \#151 ; 0*960\#171 ; { 0*951 '1-5011 \#166 ; 3-344 3-373 October 0-402 0*6801 L 0*668 J 0*671 1-494 0107 0*068 0-125 \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; L1-529 ) 1-496 December 0*407 1*505 0*127 0*097 0*154 \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; .
Mean for 1897 ... 0-410 0*664 1*493 0*132 0*090 0-088 - - \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; 0*954 1*510 3*316 1898 .
Jan nary\#151 ; Febr uary March 0*371 0*396 0*650 0*664 1*456 1*580 0*066 0*115 0*049 0*052 0*025 0*129 - * - - - = - \#151 ; 0*922 0*933 0*911 0*961 0*954 0*939 0*945 0*937 1*468 1*440 1*554 3*306 3*290 April 0*386 0*650 1*471 0*097 0*057 0*254 \#151 ; \#151 ; \#151 ; 1*501 3-326 .Time 0*391 0*663 1*649 0*061 0*001 0*220 \#151 ; \#151 ; \#151 ; \#151 ; 1*498 1*506 1*506 1*501 3-323 September 0*365 0*654 1*525 0*047 0*002 0*247 \#151 ; \#151 ; \#151 ; \#151 ; A.4A7 0*671 October 0*401 0*678 1*536 0*069 0*021 0*133 \#151 ; \#151 ; \#151 ; \#151 ; " \#151 ; * U w| November 0*371 0*658 1*514 0*062 0*025 0*187 \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; 0*659 3*307 December 0*420 0*668 1*593 0*059 -0*011 0*285 \#151 ; \#151 ; \#151 ; " 0*405 Mean for 1898 ... 0388 0*661 1*540 0*072 0*025 0*185 - - - - - - 0*406 0-665 0*944 1*497 3*310 1899 .
January 0-389 0*659 1*550 0*041 -0*010 0*330 0*628 1-019 [ 1-860 ] 0*406 0-681 0*935 1*472 r 1*502 \#166 ; t1-538 3*389 May 0*379 0*663 1*396 0*076 -0*005 0*222 3*568 4*637 8*011 0*610 1*013 2*439 0*415 0-702 0*970 V 3*331 June 0*398 0*659 1*405 0*059 -0*021 0*037 3*564 4*634 8*074 0-640 1*042 2*456 0*424 0-693 0*957 J 0*942 \0 946 1*494 j 3*265 September / 0*395 j- 0*658 1*594 / 0*0561 00*048 J l \#151 ; 0*030 0*291 f 3*568 } 4*630 7*345 / 0-660 \ 0*641 \ 1*044 2*345 0*424 0-702 11*475 3*330 10*400 1 \ 3*553 J J \#151 ; Mean for 1899 ... 0*392 0*660 1*486 0*056 -0*016 0*220 3-563 4-631 7*977 0-636 1-030 2*413 0-417 0 *694 0*950 1-490 3*329 Dr. C. Chree .
Investigations on expert .
On this clay observations were taken at the ice , steam , and sulphur points , and on the same day observations were also taken with K3 and K4 .
Further , both Mr. Hugo and myself observed , each with two sets of coils , at the ice and steam points , and our observations were in excellent agreement .
Thus I think the evidence is almost conclusive as to the reality of a fall in the Iv , , and IR of K5 , in consequence presumably of its first exposure to the sulphur point temperature .
This fall was apparently about 0o , l in the case of R0 .
This is not the only case in which this phenomenon occurred .
Thus on April 24 , 1896 , the thermometer K2 , which had been freshly remade after a breakage in silver , gave an R4 of 356-507 immediately prior to a sulphur point experiment , while on April 27 , the value found for R4 was 356-404 .
There is no reason to suppose that any sensible change in the leads had intervened .
S 43 .
Much the clearest evidence of sudden changes arises , however , in the case of exposure to the temperature of molten silver .
On December 24 , 1895 , K2 was exposed to this temperature for the first time at Kew , * the experiment being immediately preceded and succeeded by steam point observations .
It was found that IR had changed from 357-535 to 357-014 , representing a fall of 0'-52 .
On the next trial in steam the value found for R4 was 357-057 ; but as the IR of K3 and K4 showed also a slight apparent rise during this interval , the cause was most probably a slight change in the leads .
In March , 1896 , K\gt ; broke in silver and was remade .
.After exposure to the temperature of sulphur vapour on April 24 , as already described , it was tried in molten silver on April 27 ; on which occasion its IR fell from 356-404 to 356-050 .
After a second sulphur point experiment on June 26 , when no certain change occurred , K2 was tried a second time in silver on July 2 , and its IR again fell , though only by 0'*15 .
Altogether , between April 24 and July 2 , 1896 , the IR of K2 showed a total apparent fall of 0'-735 , whilst the corresponding mean apparent fall in K3 , K4 , and K5 , as obtained by interpolation , is only about 0'-015 .
The mean fall in these last three thermometers is doubtless a nearly correct measure of contemporary change in the leads ; and we thus conclude that two exposures to the sulphur point temperature and two exposures to molten silver had produced a fall of approximately 0'-7 in the zero of K2 .
The thermometer ER had been in molten silver in September , 1895 , before it came to Kew .
On its first exposure to molten silver at the Observatory\#151 ; the exposure being as usual preceded and folloAved by steam point observations\#151 ; the fall of IR was only 0'-08 ; and on two subsequent occasions the effect , if any , was very small .
For instance , on the last occasion the apparent fall was only 0'"009 .
* It was said to have been heated in September , 1895 , at Cambridge , to fully as high a temperature .
Platinum Thermometry at Kew Observatory .
47 S 44 .
We have next to consider the constancy of the fundamental intervals .
Corresponding ice and steam point observations were almost invariably taken in immediate succession to one another , so that ordinary changes in the leads could hardly affect the results .
I have divided the data into groups according to the date , and give only the mean results .
In the first group there were only two observations and in the last group only three , except in the case of Ki and K5 , where five observations had been made .
In the three intermediate groups the means are nearly all based on six observations .
A few observations made in the early part of 1897 have been wholly omitted , as the state of the leads was certainly not then satisfactory .
Table XVI.\#151 ; Mean Values of Fundamental Intervals at different Epochs .
Calibra- Epoch .
tion Kj .
k2 .
k3 .
k4 .
k5 .
Kfi .
k7 .
used .
99 + 100 + 99 + 99 + 100 + 250 + 99 + 1895 i . .
, .
.956 .711 , , .246 1896 i .812 . .
.981 .709 .247 .301 .295 Sept. , 1897 , to \ Apr. , 1898 .
J hi .776 . .
.942 .672 . .
.157 .265 Apr. , 1898 , to 1 Jan. , 1899 .
J hi .752 . .
*945 .671 . .
.163 .257 | May to Sept. , 1899 IY .712 .857 .925 .656 .188 .171 .256 | According to the table there is a fall in the fundamental interval of every thermometer between 1896 and 1899 .
It is probable , however , that this is mainly if not entirely fictitious except in the case of Ki .
When the resistance box was altered in the summer of 1897 , the coils , as Table IV shows , were slightly disturbed ; and in all probability the mean box unit was slightly altered .
An increase in this unit of about 35 parts in 100,000 would account pretty satisfactorily for the apparent differences in the results of the second and third epochs of the table .
During the third and fourth epochs nothing happened to the box , and Kx is the only case where there is any noteworthy difference between the corresponding means in the two epochs .
Between the fourth and fifth epochs the coil I broke , and its resistance was diminished by about 0*66 of a box unit by the resoldering .
This fact , and the comparatively small number of observations on which the means are based during the fifth epoch , predispose me to query the apparent falls in the fundamental intervals of K3 and K4 .
In the case of Kj , however , the fall is too large to be explained away .
S 45 .
Instead of attempting to trace changes in the ice-sulphur oo Table XVII.\#151 ; Mean values of pts[=(100/ 1 ) ( Rs - R0 ) ] for Epochs stated .
Epoch .
Using the Callendar-Griffiths formula .
Using the formula ( 17 ) .
Kx .
Ko .
K3 .
K4 .
K5 .
K7 .
K , .
K2 .
K3 .
K4 .
K5 .
K7 .
1895-6 1897 1898 1899 421 + 421 + 421 + 421 + 421 + 421 -f* .40 -39 *57 *62 - 57 \#151 ; .45 \#151 ; -54 -59 \#151 ; *47 .51 \#151 ; -56 -56 \#151 ; -49 .60 -43 -54 -58 -52 '44 421 + 421 + 421 + 421 + 421 + 421 + .35 -33 *55 -58 -51 \#151 ; *44 \#151 ; *55 -61 \#151 ; *49 .49 \#151 ; -55 -56 \#151 ; -45 .58 -46 -52 -56 -52 *43 Table X"V III.\#151 ; Mean values of 8 [ in Callendar 's formula ( 7 ) ] for Epochs stated .
Epoch .
Using the Callendar-Griffiths formula .
Using the formula ( 17 ) .
Ki- K2 .
kv k4 .
k5 .
k7 .
Kj .
K2 .
K3 .
K4 .
K5 .
K7 .
1895-6 1897 1898 1899 1-510 1-511 1-499 1-496 1 499 \#151 ; 1-507 \#151 ; 1-501 1-498 \#151 ; 1-500 1-503 \#151 ; 1-500 1-500 \#151 ; 1-504 1-497 1-508 1-501 1-498 1-502 1-508 1-513 1-515 1-500 1-499 1-503 \#151 ; 1-508 \#151 ; 1-500 1-497 \#151 ; 1-504 1-504 \#151 ; 1-500 1-500 \#151 ; 1-507 1-499 1-506 1-502 1*500 1-502 1-508 Dr. C. Chree .
Investigations on Platinum Thermometry at Kew Observatory .
49 interval R. , \#151 ; R0 directly , I pass to the consideration of and 8 , as quantities of more immediate interest .
The constancy of one of these quantities is involved in that of the other ; but each has an interest of its own .
I have thought it desirable to give the results obtained from formula ( 17 ) as well as those obtained by accepting the Callendar-Griffiths formula ( 10 ) .
The thermometer called K2 in 1896 is wholly different from that so denominated in 1899 .
With the exception of the 1899 results in K2 , and the 1897 results in Kv , every mean in the table depends on at least four observations ; in some instances eight were available .
The existence of a gradual rise in thp of Kb with corresponding fall in its S , is placed I think beyond doubt .
In the case of the other thermometers there is no fluctuation which might not reasonably be attributed to errors of observation or uncertainty in the pressure reduction formula .
Differences between the Thermometers .
S 46 .
A difference between the values of R0 or I in two thermometers does not necessarily imply any difference between them as measurers of temperature .
The temperature p is a ratio , and so should depend solely on the law of increase of resistance with temperature .
If the drawing of platinum wire affects its physical properties , we may perhaps expect differences between the temperature relations of wires of different diameters .
But there it no obvious reason a priori why differences should exist between specimens of wire of the same thickness from the same sample of platinum .
It is clearly desirable to ascertain whether such differences exist , and if they do whether they follow any recognisable law .
This is the more important at present owing to the recent proposal made by Prof. Callendar to the British Association to set up standard platinum thermometers made of some one sample of platinum wire .
Referring to Tables XVII and XVIII , we see evidence of differences between certain of the thermometers .
The consistency of the mean values found in different years for pts in K3 and K4 renders it impossible to ascribe to experimental error the very considerable differences between the values of pts in these two thermometers , and in the original K4 and K2 .
S 47 .
The evidence afforded by Tables XVII and XVIII as to differences between K3 , K4 , K5 , and Kr is less conclusive .
I have thus had recourse to two more sensitive methods of detecting differences .
The first of these consists in taking the ratios borne to one another by the resistances at the ice , steam , and sulphur points .
The taking note of the value of Ri/ Ro was in fact recommended originally by Mr. Griffiths as a delicate means of checking the accuracy of individual VOL. LXVII .
E Dr. C. Chree .
Investigations on observations , and the practice was continued for some time after the experiments commenced .
The value of these original calculations is , however , prejudiced by the fact that the results are affected by any want of equality between the leads , unless this can be eliminated .
I have thus included in the following table only those observations in which the difference between the leads was known and could be allowed for .
The thermometers are arranged in descending order of Ei/ R0 during the last epoch .
Table XIX .
Thermo- meter .
Mean values of Ri/ R0 .
Mean values of B*/ R0 .
Sept. 21 , 1897 .
( Cal .
III ) .
Oct. , 1898 , to Jan. , 1899 .
( Cal .
III ) .
May\#151 ; Sept. , 1899 .
( Cal .
IY ) .
May\#151 ; Sept. , 1899 .
( Cal .
IV ) .
*6 K , k4 Ki *7 K2 Mi 1-38771 1 -38751 1 -38707 1 -38709 1 38670 jans , excluding K 1 -38792 1 -38777 1 -38759 1 -38714 1.38706 1 -38664 1 -38787 1-38774 1-38751 1 -38711 1 -38691 1 -38664 1 -38610 2-63527 2 -63331 263165 2 -63112 2 -62937 2 -62709 1 -38702 2 -63090 The last figure is retained mainly to give an idea of the concordance between the several results for the same thermometer .
Without realising this , one is not in a position to judge of the weight to be attached to the apparent differences between the different thermometers .
It will be observed that Ki alone has altered its relative position since September , 1897 .
Again , it will be noticed that the order of R8/ R0 is the same as that of Ri/ Ro .
But for this fact we should have found larger differences between the values of g\gt ; ts and S in the several thermometers .
S 48 .
The above table will , I think , be considered pretty conclusive as to the reality of the differences between the different thermometers .
I shall , however , also give the results of the second method of attacking the problem , as they possess an independent interest .
If two platinum thermometers differ only in the absolute resistance of their spirals , the difference may be ascribed to a definite portion of the wire forming the spiral of higher resistance .
Thus , in such a case , the difference between the resistances should increase with the temperature in precisely the same way as the total resistance of Platinum Thermometry Observatory .
51 either thermometer .
Consequently , if the platinum wire in the several thermometers were identical , the differences between the resistances of any two of them at the steam point and at the sulphur point\#151 ; or the differences between any one and the mean of any two others\#151 ; should bear to the difference of the resistances in ice constant ratios .
And according to Table XIX the values of these two constant ratios should be approximately T387 : 1 and 2-631 : 1 .
Table XX shows what the ratios actually are in the cases examined .
In treating Kg I utilised only those observations in which the difference between the leads could be allowed for .
The meaning of the results will be rendered clear by an example .
The mean of the differences between corresponding observed values of Rx in K3 and K4 during 1896 was 0665 , while the mean of the differences between the observed values of Ro was 0405 : the ratio of these two quantities is T64 , and it is given under the heading Ri/ R0 .
This is here used as an abbreviation for ( Ri in K3 less Rx in K4 ) -r- ( Ro in K3 less R0 in K4 ) .
Table XX .
Epoch .
k\#187 ; - k4 .
Kj \#151 ; MK3+K4 ) .
K2-i(K3 + K4 ) .
K6-*(K .
, + K4 ) .
Rp/ Ro .
Es/ Eo .
E]/ E0 .
Es/ E0 .
Ei/ E0 .
Es/ Eo .
Ei/ E0 .
Rs/ Ro* 1896 1-64 3-78 0-89 0-53 1-61 3-65 1897 1-62 3 64 0-68 0-67 # , , .
1-61 .
\#171 ; i 1898 1-70 3-97 0-35 2-57 , , , , , . . .
1899 1-68 3-79 -0-29 3-93 1 -30 2*24 1 -62 3-79 Epoch .
lK6-i(K3 + K4 ) .
i(K3 + K4)-K7 .
k2- -K5 .
I k2- -k7 .
Eq/ E0 .
Ep/ Rq .
R\#171 ; /Rq .
Ei/ Ro- Rs/ R'o .
Ei/ R0 .
Es/ R0 1896 1-57 1897 1-63 1-58 3*48 . .
, , . .
1898 1-64 1-59 3 *51 , , , .
# , , , 1899 1 -67 1 -57 3*50 1*23 1-90 1 -36 1 2ol In only one case , K2 - K7 , are the ratios even approximately equal to the mean values given by Table XIX , viz. , T39 : 1 and 2'63 : 1 .
Even in that case the differences are too big to ascribe to experimental error , because the R0 of K2 exceeds that of K7 by fully 4-5 box units .
E 2 Dr. C. Chree .
Investigations on In the case of K3 \#151 ; K4 , and of f-Ko \#151 ; 4(K3 + K4 ) , the differences of resistance at the ice point are only about 0-4 of a box unit ; but the ratios are based on a considerable number of harmonious observations , and the smallness in their fluctuations from one year to another appears incompatible with any serious error .
The only case in Table XX in which there is unmistakable change in the values of Ri/ R0 and Rs/ R0 with the time is Kx --|(K3 + K4 ) .
The change in this case is conspicuous , and supports , of course , our previous conclusion that Kx has undergone some alteration .
S 49 .
If we omit combinations containing Kx or K2 , we notice a curious resemblance between the values of the ratios in the different cases , and this same similarity appears in other combinations of the five thermometers K3 , K4 , K5 , K~ , and K0 ( when multiplied by 0'4 ) which I have tried .
We always get something pretty close to :\#151 ; Ice point difference : steam point difference : sulphur point difference : : 1 : 1'6 : 3"6 .
The phenomena presented by these five thermometers are thus pretty much the same as if their spirals consisted in the main of given equal lengths of identical platinum , with , in addition , variable lengths of a material whose temperature coefficient is higher than that of platinum .
In this imaginary material the resistances at the ice , steam , and sulphur points would be approximately as 1 : T6 : 3'6 .
Another way of stating the fact would be that the changes in the resistance of the imaginary material due to temperature are about T56 times as great as in platinum .
In addition we can closely reproduce the figures resulting from the comparison of K2 with the other thermometers by supposing that K2 has a smaller quantity of this imaginary foreign material than the others , but a larger quantity of pure platinum .
The figures to which my calculations lead for the resistances at 0 ' , due to the foreign material and the pure platinum , are as follows , in terms of the mean box unit:\#151 ; .mometer .
Resistance , foreign material .
Resistance , platinum wire .
k2 \#163 ; # + 5*150 Kv 2 + 0-637 X k4 2+1-391 X Kb 2+1-783 X k5 2 + 2-223 X Taking , for instance , the cases K2 - K5 and K2 - K7 in 1899 , we have , in mean box units\#151 ; Platinum Thermometry at Kew Observatory .
53 Observed differences .
Calculated differences .
Ice point .
Steam point .
Sulphur point .
Ice point .
Steam point .
Sulphur point .
lO MM 1 1 Cl Cl M M 2 927 4-513 3-604 6-124 5 -564 11 -306 i ( 2 -927 ) ( 4 -513 ) 3*586 ( 6*124 ) !
5 -547 11-257 The numbers enclosed by brackets were used in my original calculation , so their coincidence with the observed values is without significance .
The coincidences could have been improved by supposing that the amounts of pure platinum in K3 , K4 , K5 , and Kr are only nearly and not absolutely the same , and by slightly altering the values 16/ 10 and 36/ 10 assumed for the ratios borne by the resistances of the foreign material at the steam and sulphur points to the resistance at the ice point .
That the above coincidences are purely accidental seems improbable , but what the real nature of the physical fact underlying the figures is , I do not know .
Accuracy of the Observations .
S 50 .
It is difficult to say what is an equitable basis on which to discuss the question .
Presumably if the room containing all the apparatus were kept at a constant temperature , and a uniform source of current were employed ; and if arrangements were made for taking the bridge centre , for reversing the battery , and for interchanging the leads and proportional arms at every observation , a variety of sources of uncertainty affecting the Kew results would be much reduced .
To provide , however , for all possible interchanges must add to the cost of platinum thermometry , and tend to make observations more tedious .
I am thus disposed to think that , except in the case of the most refined physical work , where apparatus is abundant and time and expense of relatively small account , the conditions are unlikely to be more favourable , and will in general be less favourable , than those prevailing during the experiments at Kew Observatory .
Under any circumstances a great deal may depend on whether it is , or is not , deemed essential to make a fixed point observation\#151 ; -either in ice or steam\#151 ; immediately before or after each series of observations .
If a fixed point observation be not taken , we may presumably expect uncertainties in the calculated temperatures of the same order as the fluctuations observed at Kew in Ro or Ri while if a fixed point observation is taken , we may presumably expect , when measuring a steady temperature between 0 ' and 100 ' C. , a similar degree of accuracy to that met with in the determinations of fundamental intervals in the present inquiry .
Dr. C. Chree .
Investigations on An idea of the degree of accuracy in the two cases is afforded by the two following tables .
The first gives the observed values of E0 on the days of regular observation since Calibration IY was made ; the second gives the corresponding values found for the fundamental intervals .
In the first table I give separately the values found when an allowance for the observed difference between the leads is omitted and applied .
Table XXI.\#151 ; Observed Values of Eq. Date .
Leads not interchanged .
1899 .
Kx .
k2 .
k3 .
k4 .
k5 .
k6 .
k7 .
257 + 261 + 257 + 257 + 258 + 645 + 256 + May 10 ... .
*761 .253 .874 .495 .295 .260 .715 ?
Tune 27 , 30. .
.693 .198 .833 .435 .274 .168 .677 July 10 ... .
.706 \#151 ; \#151 ; \#151 ; .280 \#151 ; \#151 ; Sep. 14 , 15 . .
.672 .184 .814 .419 .276 .132 .674 Sep. 18,19,20 .693 .198 .845 l .445 .286 \#151 ; .699 Date .
Leads interchanged .
1899 .
Kx .
k2 .
k3 .
K4 .
k5 .
k6 .
K : .
257 + 261 + 257 + 257 + 258 + 645 + 256 + May 10 ... .
.768 .260 .881 .502 .302 .267 .722 June 27 , 30. .
.709 .214 .849 .451 \#151 ; .184 July 10 ... .
.731 \#151 ; \#151 ; \#151 ; .310 \#151 ; \#151 ; Sep. 14 , 15 . .
.691 .202 .832 .436 .295 .146 .695 Sep.18,19,20 .715 .220 .86J .460 .303 1 .716 1 Table XXII.\#151 ; Observed Values of th Fundamental Interval ( leads not interchanged ) .
Date .
Kx .
k2 .
k3 .
k4 .
k5 .
1 j K6 .
| Kt .
1899 .
99 + 99 + 99 + 99 + 100 + 250+ !
99 + May 10 , 11 .712 .863 .936 .652 .197 I -203 i .262 June 27 , 30 .710 .860 .921 .660 .192 j -148 .253 July 10. . . .
.719 \#151 ; \#151 ; \#151 ; .186 \#151 ; \#151 ; Sept. 14 , 15 .701 .849 .918 .655 .171 .162 1 .254 Sept. 20. . . .
\#151 ; \#151 ; \#151 ; \#151 ; .196 | " The data in Tables XXI and XXII are a very fair sample of the Kew results .
They were taken during a time when the difference Platinum Thermometry at Kew Observatory .
55 between the leads did not vary much , so that the application of a correction for this difference does not much improve the agreement between the results of different days .
The value of E0 is conspicuously high on May 10 in all the thermometers .
This was clearly not the fault of the ice , because , as Table XXII shows , the fundamental intervals found on May 10 were not below the average , but in fact rather above .
Again , the cause can hardly have been error in the correction for temperature .
For on May 10 the temperature was only very slightly above 20 ' C. , while on the next occasion , June 27\#151 ; when the values of R0 were smaller\#151 ; it was nearly 22 ' C. \ and if we may trust Dr. Harker s experiments , the temperature coefficient applied in the reductions is certainly not too low .
Phenomena similar to those occurring on May 10 were by no means uncommon .
S 51 .
Additional information as to the degree of accuracy attained in determining the fundamental interval is afforded by Table XXIII .
This gives the mean difference , irrespective of sign , between the individual results and their arithmetic mean for two epochs , viz. , the year 1896 , and the period from September , 1897 , to January , 1899 .
In the former epoch there were six observations with each thermometer , in the latter epoch from twelve to fourteen .
Table XXIII.\#151 ; Mean Departures of Fundamental Intervals from their Arithmetic Means .
Epoch .
K , .
k3 .
k4 .
0 *4E.g .
k7 .
1896 6-015 : 6-005 0-006 6-007 6-003 o September , 1897 , to January , 1899 0*013 | 0-006 | 0-010 \#151 ; 0 -008 0-015 During the former epoch the original box with brass plug holes was in use , while during the latter epoch the plug holes were of fusible metal in Doulton ware .
During the second epoch bridge centres were taken every day , and readings were always taken with the current both ways ; in the first epoch neither of these precautions was taken .
Yet we see that the mean departures are almost all greater during the second epoch than during the first .
This certainly does not point to any superiority in fusible metal as compared to brass , but quite the reverse .
The size of the mean departures in Ki is partly due to change in the fundamental interval , especially during the second epoch .
The fundamental interval in Ko , being 2'5 times as large as in the Dr. C. Chree .
Investigations on other thermometers , has to be multiplied by 04 to render it comparable as an interval of temperature .
S 52 .
Particulars of the degree of accuracy attainable in determining pts may also be of interest .
In this case , of course , the uncertainty existing as to the true law of variation of the boiling point of sulphur with pressure should be borne in mind .
Taking all the observations made with K2 , K3 , K4 , K5 , and K7 , excluding one with K4 and the first three with K7 as probably faulty , and employing the Callendar-Griffiths formula ( 10 ) , I find for the average departure of individual values of pts from their arithmetic mean the following results:\#151 ; Table XXIY.\#151 ; Average Departure of Values of pts from their Arithmetic Mean .
k2 .
k3- K5 .
k7 .
Number of observations ... Average departure .
4 6 \#177 ; 0-04 24 \#177 ; 0-09 21 \#177 ; 0-07 10 \#177 ; 0-08 12 \#177 ; 0-07 The substitution of formula ( 17 ) for the Callendar-Griffiths formula reduces the average departures in the case of K3 , K( , and Iv7 to \#177 ; 007 , \#177 ; 0'05 , and \#177 ; 0'05 respectively , but increases the mean departures in K2 and K5 .
Roughly speaking , the average departure from the mean in the case of pts is nearly ten times as large as in the case of the fundamental interval .
This , however , may partly arise from the fact that sulphur point observations were frequently taken on the day subsequent to that on which was determined the value of R0 used in calculating pts .
Improvements suggested in Apparatus .
S 53 .
The experiments have suggested various desiderata , which it may be well to summarise .
1 .
It is desirable that some simple and more perfect method should be devised for eliminating the effects of changes in the relative resistance of the leads and the proportional arms .
2 .
The temperature of the resistance coils and bridge wire should be exposed to much less uncertainty than in the Kew box , or else these resistances should be made of some material with a much smaller temperature coefficient than platinum silver .
3 .
The bridge-wire scale is too much contracted for very exact work with thermometers of such low resistance as the ordinary Kew ones ; it is desirable , if accuracy to 0''001 C. is aimed at , that 1 cm .
of bridge wire should not answer to more than 0o , l C. Platinum Thermometry at Kew Observatory .
57 4 .
If possible , the causes of wanderings in the bridge centre and of thermo-electric currents\#151 ; not eliminated by the Griffiths key\#151 ; should be removed .
5 .
Unless the plug resistance uncertainties can be reduced , it would seem desirable to increase the resistance inside platinum thermometers intended for very exact measurement of very slowly altering temperatures , such as occur in " fixed point " observations .
6 .
Unless the heating effect of the current can be exactly ascertained and allowed for , the sensitiveness of the galvanometer should be largely increased , so as to render unnecessary currents whose heating effects will sensibly influence the readings .
7 .
It is certainly desirable that the entire room containing the apparatus should be kept approximately at the temperature accepted as the standard temperature for the coils .
At the same time it is not desirable that the observer should be continually exposed to so high a temperature as 20 ' C. A temperature such as 62 ' F. is very much healthier , at least in winter , and is more bracing .
There is no obvious reason why there might not be one standard temperature for winter and another for summer , so long as the temperature relations of the resistance coils are accurately known .
Further Experiments Wanted .
S54 .
I should like to see experiments made on the following points :\#151 ; 1 .
The true law of the variation of the boiling point of sulphur with pressure ; 2 .
The behaviour of new platinum thermometers under a variety of conditions ; 3 .
The effect of long or frequent exposures to temperatures extending from that of sulphur vapour to the highest temperature for which it is claimed that platinum thermometers are suitable ; 4 .
The effect of exposure to low temperatures , such as are now attainable by means of liquid air ; 5 .
The degree of accordance in the results deduced by Callendar 's formula from observations taken with different platinum thermometers\#151 ; varying in their \#151 ; at two or three fixed temperatures differing considerably from that of sulphur vapour , these temperatures being obtainable with a degree of accuracy not inferior to that attained in sulphur point observations ; 6 .
The cause of the curious phenomenon described in SS 48 and 49 .
S 55 .
The great majority of the observations on which this paper 58 Mr. A. E. Tutfcon .
A Comparative depends have been made by Mr. W. Hugo , senior assistant at Kew Observatory , to whose care and exactitude the research owes a very great deal .
The reduction of the observations is also mainly due to Mr. Hugo .
In several of the special experiments Dr. J. A Harker has given valuable assistance , and various operations have been much facilitated by means of apparatus devised by him .
The main work of checking the reductions and the various calculations requisite in discussing the observations have been done by myself .
I have also taken a considerable number of observations , and in discussing the various sources of trouble I speak from personal experience , supplemented of course by frequent exchange of ideas with Dr. Harker and Mr. Hugo .
At various stages of the investigation I have had valuable advice and assistance from Professor Carey Foster and Mr. W. N. Shaw , to whom the initiation of the research is largely due .
The original cost of the apparatus was partly defrayed by a grant of \#163 ; 100 from the Government Grant Fund , and a contribution of \#163 ; 120 was received last year from the Gunning Fund towards this and other cognate work undertaken by the Kew Observatory Committee .
" A Comparative Crystallographical Study of the Double Selenates of the Series R2M(Se04)2,6H20.\#151 ; Salts in which M is Zinc .
" By A. E. Tutton , B.Sc. , F.RS .
Deceived March 5 , \#151 ; Read March 15 , 1900 .
( Extended Abstract .
) In two communications to the Chemical Society , * the author presented the results of a detailed study of twenty-two salts of the series of monoclinic double sulphates R2M(S04)2,6H20 , in which R was represented by potassium , rubidium , and caesium , and M by magnesium , zinc , iron , manganese , nickel , cobalt , copper , and cadmium .
The first of these memoirs dealt with the external morphology of the crystals , and the second with their internal physical properties .
The present investigation refers to the less-known analogous double selenates , in which R is again represented by the alkali metals potassium ( atomic weight 39 ) , rubidium ( atomic weight 85-2 ) , and caesium ( atomic weight 132'7 ) .
The work on the group containing zinc has been completed , and the results are now communicated .
Topsoe and Christian-sent included in their well-known investigation the potassium salt of the group .
* ' Journ. Chem. Soc. Trans. , ' 1893 , vol. 63 , p. 337 , and 1896 , vol. 69 , p. 344 .
f 'Ann .
de Chim .
efc de Pkys .
, ' 1874 , p. 5 .

