445 An Experiment with the Balance to Find if Change of Temperature has any Effect upon Weight .
By J. H. Pointing , Sc. D. , F.R.S. , and Percy Phillips , M.Sc .
( Received July 12 , 1905 .
) In all the experiments hitherto made to determine the gravitative attraction between two masses , the temperature has not varied more than a few degrees , and there are no results which would enable us to detect with certainty any dependence of attraction upon temperature even if such dependence exists .
It is true , as Professor Hicks has pointed out , * that Baily 's results for the Mean Density of the Earth , if arranged in the order of the temperature of the apparatus when they were obtained , show a fall in value as the temperature rises .
But this is almost certainly some secondary effect , due to errors in the measurements of the apparatus , or to the seasons at which different attracted masses were used.f The ideal experiment to find if temperature has an effect on gravitation would consist in one determination of the gravitative attraction between two masses at , say 15 ' C. , and another determination at , say , the temperature of boiling liquid air .
But the difficulties of exact determination at ordinary temperatures are not yet overcome , and at any very high or very low temperatures , they would be so much increased that the research seems at present hopeless .
The question can , however , be attacked in a somewhat less direct method by examining whether the weight of a body\#151 ; the gravitative attraction of the earth upon it\#151 ; varies when the temperature of the body varies .
The various parts of one of the attracting masses\#151 ; the Earth\#151 ; remain , each part , at the same temperature throughout , and this is , no doubt , a weakness of the method .
For it is perhaps conceivable that in the expression for the attraction a temperature factor might exist of some such form as 1 + + + m ) , where M and m are the two masses , and t and t ' are their temperatures .
If m/ M is negligible , this reduces to 1 + and is independent of the temperature t ' of the smaller mass .
But it seems more likely that each mass would have a separate temperature factor .
If such a factor exists , and if its variation is appreciable , then we ought to be able to detect a change of weight with change of temperature .
Observations on pendulums suffice to show that at the most any such effect * 'Proc .
Camb .
Phil. Soc. , ' vol. 5 , p. 156 .
t Pointing , " Mean Density of the Earth , " p. 56 .
Dr. J. H. Pointing and Mr. P. Phillips .
[ July 12 , must be small .
The nearly constant period of vibration with the nearly constant length of a compensated or an " invar " pendulum shows constancy of weight of the bob to a considerable degree of exactness .
Again , the agreement of weight methods and volume methods of measuring the expansion of liquids with rise of temperature show , though less conclusively , that there is no great variation .
It appeared to us that it would be possible to go much further in testing constancy of weight by a direct weighing experiment , in which the weight on one side of a balance should be subjected to great changes of temperature while the counterpoise should remain at a uniform temperature .
We give an account in this paper of a series of experiments carried out on the following principle .
A brass cylinder weighing 266 grammes was hung by a wire from one arm of a balance so as to be near the bottom of a tube depending from the floor of the balance case , the tube being closed at the bottom and opening at the top into the case , the wire passing down through the opening .
The brass cylinder was counterpoised by an equal cylinder hung by a short wire from the other arm inside the case .
To this short wire was attached a finely divided scale on which the swings of the balance could be read by a microscope looking through a window in the case .
The balance was released and left free to swing .
Then the case was exhausted till the pressure was not more than a small fraction of a millimetre of mercury .
Steam was passed round the lower part of the tube where the weight hung , and after a time the weight was allowed to cool again .
In other experiments the lower part of the tube was cooled by liquid air and again brought up to the temperature of the room .
While the changes of temperature were in progress there were considerable apparent variations in weight .
But ultimately , when the temperature became steady , the weight , too , became steady .
At 100 ' C. it was slightly less than at the temperature of the room .
This difference was partly due to a rise in the temperature of the case , such a rise being always accompanied by an apparent diminution of the weight in the tube , whether steam was applied or the balance was merely left to follow the temperature of the room .
Probably this effect was due to some change in the balance beam .
But the difference was partly due to convection currents , or at any rate to the residual air in the case , for it varied with the disposition of diaphragms in the tube .
There were no doubt convection currents , as there was always a tendency for the case to rise in temperature when steam was applied , and this could hardly be accounted for by conduction or radiation , under the conditions of the apparatus .
As effects due to residual air should depend upon surface and not upon volume , similar experiments were made with hollow weights , each 1905 .
] Experiment to find if Temperature effects Weight .
447 about 58 grammes , and of the same size and form as the solid weights .
There was again an apparent diminution in weight when steam was applied .
Any true diminution due to change of temperature should be shown by a difference in the diminution with the solid and with the hollow weights , the surface effects being eliminated , and this diminution should be that of 266 \#151 ; 58 = 208 grammes .
The net result of all the experiments was that there was not a greater change in 208 grammes between 15 ' C. and 100 ' C. , than 0-003 milligramme .
But an inspection of the detailed account given later shows that this result is probably accidentally small\#151 ; within the limits of experimental error .
It would imply that there is not a change greater than 1 in 6 x 109 per 1 ' C. But the experiments hardly justify us in saying more than that there is not a change greater that 1 in 109 per 1 ' .
When liquid air was used , air currents were absent , and the temperature variations of the case were much less .
The net result of these experiments was that there is not a change of weight in 208 grammes between 16 ' C. and \#151 ; 186 ' C. greater than 0*002 milligramme .
This would imply that there is not a change greater than 1 in 1*3 x 1010 per 1 ' change of temperature .
We may probably assert that the change is not greater than 1 in 1010 per 1 ' C. We now proceed to a detailed account of the apparatus and of the mode of using it .
The Balance .
The balance has a 6-inch beam and was specially constructed for the experiment by Mr. Oertling .
The general arrangement will be seen from fig. 1 .
The base plate is of gunmetal , as are also two sides and the top of the case .
The front and back of the case are of thick plate glass fixed to the metal by marine glue .
In the experiments the base plate was supported on levelling screws on a slate slab , and between it and the slab was a gas pipe with pinhole burners so that it could be warmed .
When the case was to be fixed in position the jets were lighted and sealing wax was smeared on to the area of contact of plate and case .
When the wax was quite liquid the case was put down on the plate and the gas was turned off .
When the metal was cool the joint was perfectly air-tight .
The tube T in which the weight W hung was of brass , 4*1 cm .
internal diameter and 62*5 cm .
long .
It consisted of three parts .
The topmost was brazed to the base plate and the two lower parts were attached to it and to each other by flanged joints ff .
Between the flanges was placed a circular lead washer of diamond shaped section .
When the flanges were pressed together by bolts the joint was quite air-tight .
Round the middle section of Dr. J. H. Pointing and Mr. P. Phillips .
[ July 12 , the tube was a water jacket wjthrough which water flowed while an experiment was in progress , and round the lowest section was a steam jacket sj through which either water or steam could be passed .
This jacket could be removed and could be replaced by a vacuum vessel 30 cm .
lung containing liquid air .
W , weight of which the temperature is to be raised , W ' counterpoise .
T , tube in which it hangs , with a number of diaphragms with ^-inch holes .
sj , steam jacket , replaced by liquid air jacket .
ff , flanged joint with lead washer .
wj , water jacket .
p , pipe to the exhausting pump .
sc , scale read by a microscope not shown .
rr , rider rod passing through stuffing boxes , sb , enlarged in fig. 2 .
gg , gas burners to heat the base plate before sealing up .
The weights W W ' were turned from the same gunmetal bar .
The length of each was 4*45 cm .
the diameter 3 cm .
and the solid weights were each 266T7 grammes , while the hollow ones were 57*86 grammes .
They were hung directly from the end plates of the balance by platinum wires , and any residual inequality was compensated by moving a centigramme rider along the 1905 .
] Experiment to find if Temperature affects Weight .
449 beam by the rider rod rr .
This rod passed through stuffing boxes designed for us by Mr. G- .
0 .
Harrison , the mechanical assistant in the laboratory , to whom we are much indebted for this and many other valuable suggestions , and for the careful construction of all the apparatus except the balance .
These stuffing boxes were perfectly air-tight when screwed up and the rod could still be rotated without any leakage .
But to draw it in or out it was necessary to loosen the screws slightly , and in one case when this was done some leakage occurred .
As the construction appears to give an efficient mode of moving apparatus inside a vacuum from without , we give in fig. 2 a section of a stuffing box .
c c rr , rider rod .
CC , side of case .
ww , two or three circular washers punched out of soft leather and soaked in oil .
p , plunger driven in by screws , ss .
oh , oil hole through which valvoline , a thick lubricating oil , was inserted .
The position of the balance beam was read by a microscope viewing a glass scale sc , fig. 1 , interposed in the suspension of W ' .
The scale was divided to OT mm. and numbered in millimetres .
The objective of the microscope was placed inside the case and the eye-piece with cross hairs was fixed outside it .
The axis of the microscope was horizontal and a lamp at the back illuminated the scale .
The case was surrounded by felt and a tin cover was placed over the whole , small windows through the felt then allowing the scale to be seen .
A thermometer placed between the felt and the case was taken to give the temperature of the case .
A brass pipe p from the floor of the case led to the pumping apparatus .
This pipe was connected to a branched glass tube , one branch going to a Fleuss pump and the other to a 4-fall Sprengel , made continuous in its action by a steel pump which was worked by a motor , and which raised the mercury Dr. J. H. Pointing and Mr. P. Phillips .
[ July 12 , again from the cistern at the base to the reservoir at the top .
When the case was to be exhausted the Fleuss pump was first used and then sealed off and the exhaustion was carried on by the Sprengel .
The degree of exhaustion was estimated by sending a discharge through a vacuum bulb 10 cm .
diameter connected with the tube to the pump , and usually the pumping was continued till the negative dark space was of the order of 4 cm .
As a rule the vacuum held without serious change for days or even for weeks .
Mode of Experiment .
A large number of preliminary experiments were made with a pair of brass weights each about 187 grammes .
These were only useful in bringing to the front the difficulties in obtaining good results and in suggesting means for overcoming them .
We shall only record the final results with the 266 grammes and 58 grammes weights .
The weights and the lower section of the tube were first cleaned by boiling in caustic potash solution and washing in distilled water .
They were then suspended , being handled with gloves only , and the lowest section of the tube was screwed on .
Steam Heating .
The jacket sj(fig .
1 ) was fixed on the lower section of the tube and the balance was set free to vibrate , being left free during a whole series of experiments .
The case was then sealed on and the value of a scale division was determined by the rider .
Any change in the value during a series could be determined from the change in period of the swing .
The time of swing in different series ranged from 24 to 42 seconds .
After the stuffing boxes were tightened the case was exhausted till the pressure was estimated to be not more than'mm .
of mercury .
The weight of air displaced by a weight was then of the order CHHJl milligramme and the change in this with change of temperature was quite negligible .
Cold water was passed through the water jacket and sometimes , while steam was being got up in a boiler at some distance well screened from the balance , through sj also .
The centre of swing and the temperature of the case were observed , and before any heating occurred the balance was usually quite steady .
Steam was then blown through water still flowing through After considerable changes , which will be described later , the centre of swing in the course of five or six hours settled down to a steady march which appeared to correspond to change in temperature of the case .
Sometimes steam was turned off after eight or nine hours , but in some cases it was kept on for 24 and 48 hours and even longer .
Then it was turned off and the 1905 .
] Experiment to find if Temperature affects Weight .
451 jacket was allowed to cool .
The centre of swing was observed when steady , several hours later or next day .
The results in the first few heatings and coolings after an exhaustion of the case were rejected , as there was evidence that the earlier heatings drove gas from the weight .
Only after successive heatings gave fairly consistent values were these taken into account .
One effect of the steam heating was always to raise the temperature of the case , probably through convection of the residual air .
A rise of temperature in the case was always accompanied by a lowering of the scale reading , corresponding to a diminution in weight .
The effect was somewhat irregular , but an average value of the lowering per 1 ' rise was determined by observing the centre of swing of the balance at intervals through several days , when the balance was left to follow the varying temperature of the room and no steam was flowing .
The value thus obtained was used to correct all the readings to 15 ' C. As an example of the method pursued , we give in Table I the series of readings used to obtain the temperature correction for the hollow weights .
Table I.\#151 ; Change of Centre of Swing with Change of Temperature of Case .
Date .
Time .
Centre of swing in millimetres of scale .
Temperature of case .
6.12.04 1.0 P.M. 14 *34 14 -75 ) ) 3.0 " 14 *355 14 -55 ) ) 5.40 " 14 *35 14-65 7.12.04 12.35 " 14 *51 12 -70 1 ) 5.5 " 14 *465 13-60 8.12.04 11.5 A.M. 14 *43 14-70 )\gt ; 12.55 p.m. 14 *20 14-95 9.12.04 9.51 A.M. 14 *555 12-80 99 ... ... 2.27 p.m. 14 *525 13 -3 10.12.04 9.37 a.m. 14 *46 15 -0 The temperature correction deduced from these numbers by the method of least squares is a decrease of 0T3 division per 1 ' C. rise , and as the sensibility was 1 division for 0*248 milligramme , there was an apparent decrease of weight of 0*032 milligramme per 1 ' rise .
Two similar series with the solid weights gave a decrease of 0*044 division per 1 ' C. rise , and as the * sensibility was now 1 division per 0 803 milligramme , there was an apparent decrease of weight of 0*035 milligramme per 1 ' rise .
Another series with the solid we ; gbfs when steam was passing all the time for several days , gave a decrease of 0*052 division per 1 ' rise , but as the VOL. lxxvi.\#151 ; a. 2 i Dr. J. H. Pointing and Mr. P. Phillips .
[ July 12 , values were more irregular , the series giving 0-044 division were used .
This series with steam sufficed to show that very nearly the same temperature correction applied when the weight was hot as when it was cold .
The irregularity of the observations is only to he expected when it is remembered that the balance was subjected to some considerable vibration at times through machinery running in the same building , and that the observations extended over several days .
Indeed it is remarkable that there was not more irregularity , and the fair consistency of the observation illustrates once more the marvellous accuracy of a well-made balance .
The following Table II will serve as an example of a complete experiment in which one of the hollow weights was cold initially , was then surrounded with steam for 24 hours , and was then allowed to get cold again .
The observations recorded are at about hourly intervals , but intermediate ones , not used , were frequently taken to be sure that there were no sudden changes .
Table II.\#151 ; Experiment with Hollow Weight raised to 100 ' C. and then cooled , 1 mm. = 0'248 milligramme .
Correction for temperature of case \#151 ; 0-13 division per 1 ' .
Date .
Time .
Condition of weight .
Centre of swing , 1 = 1 mm. Temperature of case .
Centre of swing corrected to 15 ' C. Remarks .
17.11.04 9.25 a.m. Cold 14 *905 14 ' *75 14 -872 Steam put on just after )\gt ; 4.0 p.m. Hot 14 *40 15 *0 14 -400 9.25 and kept on till JJ 5.20 " yy 14 *39 15 *0 14 -390 10 A.M. next day .
M 6.12 " yy 14 *375 15 *0 14 '375 V 7.8 " yy 14 *365 15 *05 14 -372 J ) 7.56 " yy 14 *36 15 *1 14 -373 18.11.04 .
9.5 A.M. yy 14 *30 15 *05 14 -307 yy 9.55 " yy 14 *295 15 *1 14 -308 Steam turned off just yy 5.36 p.m. Cold 14 *55 16 *00 14 -680 after 9.55 .
yy 6.39 , , yy 14 *56 15 *95 14 -684 yy 7.56 " yy 14 *58 15 *75 14 -678 19.11.04 9.40 a.m. yy 14 *67 14 -30 14 -579 31 11.38 " yy 14 *665 14 *20 14 -561 Initial reading , cold at 15 ' ... . .
14*872 divisions .
Final mean reading , cold at 15 ' ... 14*636 " Mean reading , cold ... ... ... ... ... . .
14*754 " " hot ... ... ... ... ... ... ... ... 14*360 Cold\#151 ; hot ... ... ... ... ... ... ... ... ... .
0*394 division .
The following Table III gives the results of the various experiments with the hollow weight , treated as in Table II , the readings of the centre of swing being at about hourly intervals when on the same day .
1905 .
] Experiment to find if Temperature affects Weight .
453 Table III.\#151 ; Experiments with Hollow Weight raised to 100 ' C. and then cooled , 1 mm. = 0'248 milligramme .
Correction for temperature of case \#151 ; 0-044 division per 1 ' .
Date .
Condition of weight .
Centre of swing corrected to 15 ' .
Number of readings from which centre of swing is found .
Greatest deviation from the mean .
Excess of cold above hot .
Remarks .
16.11.04 Hot 14 -749 4 0 -0201 0 *123 Temperatures 14'*75 to 16'*3 .
17.11.04 Cold 14 -872 1 - J The initial cold reading was rendered useless by a subsequent shift of scale reading , probably due to slight displacement of the eye-piece .
17.11.04 Cold 14 -872 1 - 1 Temperatures 14'*2 to 16 ' .
17\#151 ; 18.11.04 Hot 14 -360 7 0 -058 ^ 0-394 The set given in full in 18\#151 ; 19.11.04 Cold 14 -636 5 0 -075 J Table II .
The last of the preceding used as the first of this .
21.11.04 Cold 14 '218 1 - 1 Temperatures 10'*45 to 14'*8 .
22\#151 ; 23.11.04 Hot 14 -091 6 0 -095 V 0-165 23.11.04 Cold 14 -294 3 0-040 j 25.11.04 Cold 14 -421 1 \#151 ; Temperatures 10'*95 to 14'*5 .
\gt ; J Hot 14 -029 2 o-ooi 0-187 26.11.04 Cold 14 -010 % 2 0-016 1.12.04 Cold 14 '500 1 \#151 ; Temperatures 16 ' to 16'*9 .
\gt ; \gt ; ... ... Hot 14-100 3 0*007 0-334 2.12.04 Cold 14 -367 1 \#151 ; 2.12.04 Cold 14 -367 1 \#151 ; Temperatures 16'*9 to 17'*45 .
2\#151 ; 3.12.04 Hot 14 -218 3 0*048 0-166 3.12.04 Cold 14 -400 1 \#151 ; 12.12.04 Cold 14 -379 1 \#151 ; Temperatures 13'*65 to 15'*75 .
... ... Hot 14 -106 3 * 0*018 0-279 13.12.04 Cold 14 -390 1 \#151 ; Mean value cold\#151 ; hot = 0*235 division = 0*058 milligramme .
The following Table IV gives the results with the solid weight .
They are not so consistent as those with the hollow weight , probably because they were spread over a longer time on the average .
This was done to secure that the weight should be more nearly at the temperature of its surroundings .
A rough estimate shows that if heat be gained by radiation alone and the brass is taken as a full radiator , three hours will be required to bring it within 1 ' of the temperature of the steam .
The last two experiments were incomplete in that no final cold weighing was taken , but the results obtained were regarded as probably sufficient .
Dr. J. H. Pointing and Mr. P. Phillips .
[ July 12 , Table IY.\#151 ; Experiments with Solid Weight raised to 100 ' and then cooled 1 mm. = 0-803 division .
Correction for temperature of case \#151 ; 0-044 division per 1 ' .
Date .
Condition of weight .
Centre of swing corrected to 15 ' .
Number of readings from which centre of swing is found .
Greatest deviation from the mean .
Excess of cold over hot .
Remarks .
26.12.04 Cold 16 *295 1 Temperatures 9'*9 to no,0 .
27\#151 ; 28.12.04 29.12.04 Hot Cold 16 -032 16 -046 2 1 0*017 0*139 30.12.04 Cold 16 *016 1 \#151 ; Temperatures 9'*1 to 12'*05 .
30\#151 ; 31.12.04 2.1.05 Hot Cold 15 *914 16 *035 4 1 0 *021 0-112 2.1.05 Cold 16 -035 1 \#151 ; Temperatures 9'T to 12'*1 .
3.1.05 Hot 16 -022 2 0*018 0-019 4.1.05 Cold 16 -047 1 \#151 ; 5.1.05 Cold 15 -987 1 \#151 ; Temperatures 11'*95 to 12 ' 65 .
Hot 15 -919 3 0*002 0-068 Experiment interrupted by stoppage of steam tubes .
9.1.05 Cold 16 -048 1 \#151 ; Temperatures 13'*7 to 15'*65 .
10.1.05 Hot 16 -039 4 0-020 0-009 Heating continued several days after this to obtain temperature correction .
No Jinal cold reading taken .
Mean value cold\#151 ; hot = 0*069 division = 0*055 milligramme .
From Tables III and IV we have\#151 ; Solid weight , 266 grammes : cold\#151 ; hot ... ... .
= 0*055 milligramme .
Hollow weight , 58 grammes : cold\#151 ; hot ... = 0-058 " For the difference , 208 grammes : hot\#151 ; cold = 0-003 " Taking the rise in temperature as 85 ' , this gives a change of the order of 1 in 6 x 109 per 1 ' rise .
But evidently the smallness of the result is accidental , and probably all we can assert from the work is that any change of weight with change of temperature between 15 ' C. and 100 ' C. is not greater than 1 in 109 .
Cooling with Liquid Air .
Experiments were made in which heating by steam was replaced by cooling with liquid air .
This was supplied to us by Sir William Bamsay , and we desire to express our hearty thanks to him for his ready kindness in helping us to increase the temperature range so considerably .
In these experiments the steam jacket was removed and replaced by a vacuum vessel 0 cm .
deep and 6 cm .
inside diameter , kept full of liquid air .
After the 1905 .
] Experiment to find if Temperature affects Weight .
455 steady state was reached the liquid air was removed , the jacket was replaced and cold water was again passed round the tube .
Owing to the evaporation of the air the experiments had to be carried out more rapidly than those with steam , but through the absence of convection currents , a steady state was more rapidly reached , and the variation in the temperature of the case was very small .
The temperature correction was not observed , but as in the subsequent observations with both solid and hollow weights , it was found to be about 0-03 milligramme per 1 ' , this value was assumed to hold here .
In any case its effect is very small , as the temperature varied so little .
The centre of swing was observed nearly continuously from the time when the liquid air was applied and again after it was removed .
After a time in each case it became steady and only these steady values are recorded in the following Tables .
Table Y.\#151 ; Experiment with Solid Weight cooled by Liquid Air , 1 mm. = 0'315 milligramme .
Correction for temperature of case \#151 ; 0T division per 1 ' .
Date .
Time .
Condition of weight .
Centre of swing .
Temperature of case .
Centre of swing corrected to 16'*6 .
Remarks .
28.7.04 ... 3.25 p.m. Normal 11 -085 16c -6 11 -085 Liquid air applied just after 3.25 .
33 5.50 " Cold 11 -07 16 -651 | ) ) 6.0 " 33 11 -07 16 -65 11 -075 33 6.10 " 33 11-07 16 -65 J 1 Liquid air removed 33 * * * 8.45 " Normal 11 -095 16 '4 ] 1 11 -075 and water applied \gt ; 3 ... 915 " 33 11 -095 16 *4 J r just after 6.10 .
Normal\#151 ; cold = 0*005 division = 0 0016 milligramme .
Table VI.\#151 ; Experiment with Hollow Weight cooled by Liquid Air , 1 mm. = 0'343 milligramme .
Correction for temperature of case \#151 ; OT division per 1 ' .
Date .
Time .
Condition of weight .
Centre of swing .
Temperature of case .
Centre of swing corrected to 16'*6 .
Remarks .
9.9.04 ... 9.40 A.M. Normal 14 *485 16'-3 14 -455 Liquid air applied at 33 11.40 " Cold 14 *480 16 -4 \ 14 -460 9.43 .
33 11.50 " 33 14 *480 16 -4 / Removed at 11.52 .
33 5.25 p.m. Normal 14 *480 16 -4 14 -460 Steady .
The balance next morning read 14*48 at 16'*3 .
Normal\#151 ; cold = \#151 ; 0*002 division = \#151 ; 0*0007 milligramme Dr. J. H. Pointing and Mr. P. Phillips .
[ July 12 , From Tables V and VI we have\#151 ; Solid weight , 266 grammes : normal\#151 ; cold ... = 0-0016 milligramme .
Hollow weight , 58 grammes : normal\#151 ; cold ... =\#151 ; 0-0007 " For the difference , 208 grammes : normal\#151 ; cold = 0-002 " Taking the fall in temperature as 200 ' , this gives a change of the order of 1 in 2 x 1010 per 1 ' fall .
These liquid air experiments were not repeated .
But the conditions are probably much less disturbed than with the steam experiments , and we may .
safely say that if there is any change of weight with change of temperature between 16'*6 C. and \#151 ; 186 ' C. , it is not so great as 1 in 1010 per 1 ' C. Note on the Change of Apparent Weight on First Heating or Cooling .
We have mentioned that while the changes in the temperature of the weight were in progress there were considerable apparent variations in weight .
These , in a few cases , amounted to as much as 0-6 milligramme .
They were almost certainly due to radiometric forces or to other gas action , for they were very dependent on the disposition of the diaphragms in the tube T ( fig. 1 ) , and also on the way in which the steam was blown through the jacket .
In the preliminary experiments with solid weights the lowest diaphragm was 5 to 6 inches above the weight , and the steam was blown into the top of the jacket .
Under these circumstances the following variations occurred when the steam was turned on :\#151 ; At first the weight apparently increased , until in 15 to 20 minutes it reached a maximum , which was in some cases as much as 0'6 milligramme above the real weight .
After reaching this maximum the weight apparently decreased , till in four hours it had reached a nearly steady value , which was a little less than the value at the temperature of the laboratory .
If , now , the jacket was filled with cold water , the apparent weight first increased for about one minute and then decreased for about two hours to a minimum , which was a little lower than the final weight at 100 ' .
After this the weight very slowly increased , till in five to six hours it had recovered the value which it had before the experiment .
These changes did not vary very much with the pressure , but at lower pressures they took place more rapidly than at higher ones .
On cooling the weight with liquid air changes occurred exactly similar to those which occurred when the weight was cooled from 100 ' to the temperature of the laboratory ; and when the weight was warmed up from the temperature of liquid air to the temperature of the laboratory , the 1905 .
] Experiment to find if Temperature affects Weight .
45 7 changes were similar to those when the weight was warmed from the temperature of the laboratory to 100 ' C. So long as the arrangement of the diaphragms and the weight remained the same , and so long as the steam was blown through in the same way , these changes were exactly similar , but as soon as any alteration was made in these arrangements the character of the changes altered .
In one series of experiments a sealed glass bulb containing mercury was used in place of the brass weight .
In this case , immediately after the steam was turned on there was a rapid decrease in weight , and a minimum was reached in less than one minute .
After this the changes were very similar to those occurring with the brass weight .
On cooling , however , the changes were almost exactly the reverse of the changes on heating , and were not at all like the changes with the brass weight .
In the final experiments , those recorded , the lowest diaphragm was within ^-inch of the top of the brass weight .
With this arrangement and with the steam blown into the top of the jacket , the following changes occurred:\#151 ; The apparent weight first increased rapidly , reaching a maximum in about one minute , then it rapidly decreased , reaching a minimum in about four minutes , and again increased to another and lower maximum in 8 to 10 minutes .
After this it slowly decreased to a nearly steady value a little lower than the original value .
On passing cold water through the jacket the apparent weight rapidly increased for about one minute , and then slowly decreased to its original value .
Still another variation was arranged by blowing the steam in at the bottom of the jacket instead of at the top , all the other things remaining as in the last experiment .
In this case , on turning on the steam , the apparent weight first decreased to a minimum in about one minute , then increased to a maximum in about six minutes , and finally decreased slowly to a nearly steady value a little below the original value .
The cooling and the consequent changes were exactly similar to those in the last experiment .
It is somewhat difficult to follow out exactly the changes which would be caused by radiometer action and by convection currents in these different arrangements of the apparatus , but the fact that these changes depend entirely on the arrangement , is sufficient evidence that they are caused by gas action , and , as we have before said , we have some reason to believe that even the small final difference is due to air currents .

