392 Experimental Investigation as to Dependence of Gravity on Temperature .
By L. Southerns , Whitworth Scholar .
( Communicated by Professor W. M. Hicks , F.R.S. Received September 13 , \#151 ; Read November 8 , 1906 .
) Few experiments have been made with a view of detecting with accuracy any alteration in the weight of a given substance which may accompany variation of its temperature .
The recent experiments of Professor Pointing and Mr. Phillips described in the 'Proceedings of the Royal Society , ' September 13th , 1905 , led to a null result.* At the time of the publication of this result the Author had been engaged for about a year on a similar line of research , and as the two methods of experiment were entirely different , it was thought advisable to continue with this work in order that the results obtained by the two methods might be compared .
Further reference will be made to this matter in the sequel ; but it may be noted here that in their experiments a mass of gunmetal was heated externally , while in the present ones a mass of paraffin oil was heated internally .
Original Apparatus and Method of Experiment .
The apparatus in its original form was constructed several years ago by Dr. Hicks .
It consisted of a specially made balance , carrying at one end of the beam a calorimeter , and at the other end a magnetic counterpoise .
The calorimeter consisted of a light aluminium vessel containing refined paraffin oil , in which was immersed a coil of fine platinum wire of about 1100 ohms resistance , wound on a mica frame .
Two other aluminium vessels surrounded the inner one , and the spaces between them were packed with cotton wool ; the vessels themselves were made airtight by sealing up with wax .
Leads from the coil passed up a tube to the suspension pieces of the calorimeter .
A magnet was suspended from the other end of the beam as counterpoise , so that by passing a small current through a coil fixed in the balance case the magnet could be attracted or repelled , and the balance thus brought to its zero position and calibrated while within its closed case .
The beam consisted of two parallel brass arms fastened together by means of a block of insulating material .
The knife-edges were of hardened steel , * It is interesting to remember that Count Rumford attempted this experiment , and concluded from its negative results that heat could not consist of a ponderable fluid ( of .
' Life and Works of Count Rumford , ' vol. 3 , p. 1 ) .
Investigation as to Dependence of Gravity on Temperature .
393 and they rested on steel plates which were connected to binding screws outside the balance case .
Each knife-edge was electrically connected to the centre of one of the parallel brass arms of the beam ; thus a current could pass from one binding screw through the knife-edge and one conducting arm of the beam , thence by a steel point to the suspension piece of the calorimeter , through the heating coil immersed in the mass of oil , and back again through the other part of the beam and its knife-edge to the other binding screw .
The heating coil could also be used as a platinum thermometer for ascertaining the temperature of the oil before and after the passage of the heating current .
A mirror fixed to the beam enabled observations to be made from a distance by means of telescope and scale .
The advantages of this design of calorimeter and method of heating are that the heating is very rapid ; that it takes place while the balance is swinging on its knife-edges ; and that the outer envelope of the calorimeter is not heated , and only slowly becomes warm , so that air currents , etc. , are reduced to a minimum .
Also , the complete experiment can be conducted from a distance by the observer .
A number of experiments were made with this apparatus , using a heating current of about 0'2 ampere , alternating .
The diagram A shows the results of a typical experiment plotted in the form of a curve , the abscissae of which represent times and the ordinates scale readings , deflections of the balance beam .
The length of the vertical line A B in the diagram represents an increase of weight of 1 milligramme , the weight of oil heated being about 250 grammes .
During the period C D observations were taken previous to heating , the balance being left free to swing on its knife-edges .
D E represents the period of heating ( one minute in this case ) , and E F part of the period of cooling , during which observations were also taken .
In all these experiments a large apparent increase of weight took place very soon after the period of heating , although the changes during the heating were very small .
In order to investigate the cause of this , a platinum thermometer was inserted in the balance case near the calorimeter , and its temperature observed from time to time during an experiment .
Diagram B shows the temperatures indicated plotted to the same time base as the balance deflections .
The ordinates of the temperature curve ( shown dotted ) are proportional to the temperatures indicated , and approximately represent in magnitude the diminution in the buoyancy of the air , supposing the air surrounding the calorimeter to be at the temperature indicated by the thermometer .
This is , of course , only a rough approximation to the truth , but the general similarity of the curves indicates that the effect is largely due to diminution of buoyancy of the air caused by escape of heat from the Mr. L. Southerns .
Investigation as to [ Sept. 13 } 4-0 TIME 12-30 1906 .
] Dependence of Gravity on Temperature .
calorimeter .
As the temperature of the oil was only raised four or five degrees in these experiments , it was obvious that this effect must be greatly reduced before trustworthy observations could be made of any change in weight which might occur during the heating period if considerable increments of temperature were to be used .
It was then decided to make the weighings in a partial vacuum in order to reduce this buoyancy effect .
Modifications of the Apparatus .
After some preliminary trials with a temporary vacuum chest , the apparatus was rearranged and set up in a constant temperature room in the basement of the new Physical Laboratory at Sheffield .
This room was for the purposes of this experiment divided into two portions by a partition , the balance being set up on one side and observations taken from the other .
The general arrangement of the apparatus is shown in fig. 1 .
In the figure A A represents the partition , to which are fixed various instruments , resistances , etc. , not indicated in the figure .
B is a brick pillar , which has a rubber course at its base , resting on a concrete slab C. The balance Mr. L. Southerns .
Investigation as to [ Sept. 13 , is placed on a cast-iron tray D D , and over it is lowered a cast-iron cover E provided with circular windows .
The bearing surfaces of cover and tray are faced with rubber , and the joint is rendered airtight by pouring mercury around it .
An enlarged section of this joint , which can he made and broken with the greatest ease , is given in the figure .
F shows a portion of the cover with its strengthening flange Gr , H the tray , J the mercury space , and K the rubber facings .
Keturning to the main figure : L L represents a trestle from which pulleys are slung for raising the cover E. M is a tube to which are connected a water pump and a Fleuss air-pump for exhausting the balance case , and also a gauge which indicates the stage of exhaustion .
Copper wires , sealed into glass tubes , which are in turn sealed into holes through the tray D , form the electrical connections for conveying the current into the balance case ; one of these is shown at N. P is the observing telescope fixed to a stand Q set up inside the portion of the room cut off by the partition so as to be free from sudden variations of temperature , vibration , etc. The scale is placed at S S , and is illuminated by means of a lamp T , with lens U , and mirror Y. Many instruments and practical details are left out of the figure in order to avoid complication .
' It was also necessary to construct a new calorimeter , as the original one was not designed to support the internal pressure produced when the balance case is exhausted .
This is shown in fig. 2 , which will be described in due course .
Also a closed vessel having a volume equal to that of the inner vessel of the calorimeter was attached to the magnetic counterpoise .
The new apparatus was first tested at full atmospheric pressure , a current of 035 ampere being used for heating .
It was found that motions of the balance were produced by the passage of the current ; thus deflections of 03 to 2-5 mm. were observed while the current was passing , but the balance nearly resumed its original position when the current was cut off .
About 60 experiments under varied conditions were made in order to ascertain the causes of these disturbances .
They were found to be due to induced currents , chiefly in the magnetic counterpoise coil , to self-induction in the beam and other parts of the circuit inside the balance case , and to some source of error due to the passage of the current through the knife-edges .
It was , therefore , necessary to make further alterations in the apparatus .
The magnetic counterpoise had to be sacrificed , and the current was made to pass by two stout copper wires fixed near each other to the beam , instead of by the brass arms of the beam , which enclosed a considerable area .
These wires were pointed and made to just dip into two mercury cups attached to the supports of the balance so that the points of contact of wire and mercury were in the same straight line as the central knife-edges .
The 1906 .
] Dependence of Gravity Temperature .
L__________J Fig. 2 .
A A. Copper calorimeter , 9 cm .
x 8 cm .
, made in two halves soldered together , containing oil to about the level shown .
Weight of the copper = 53 grammes ; of the oil 233 grammes ; of the total heated mass , including solder and coil , about 300 grammes .
\#163 ; B. Brass tube soldered to top of calorimeter .
The leads from coil pass up this tube , which is sealed airtight with khotinski cement .
C C. The dotted lines indicate shape of frame on which the fine platinum wire coil is wound ( non-inductively ) .
The wire is distributed fairly uniformly throughout the mass of the oil .
D D. Aluminium vessels packed with cotton wool .
The outer one is airtight , but the inner one is not .
E. Ebonite tube , closed at top , which retards loss of heat from B B , and carries the suspension plates .
F. Suspension plate , with glass bearing slip underneath .
One lead from coil is attached to F. G. Tinfoil strip connecting F to the conducting rod for conveying current to the coil .
H. Conducting rod fixed rigidly to beam by an ebonite clip not shown .
J. Mercury cup attached to balauce case into which the pointed end of H dips to make contact with external circuit .
The point is in same straight line as knife-edges .
K. Part of beam with bearing point .
F , G , H , J , K are in duplicate .
L. Gauge ( not shown in its true position ) containing mercury .
This allows air to be drawn from the spaces between ADD when the balance case is exhausted , but the thread of mercury falls back and prevents escape of residual air from calorimeter during an actual experiment .
Mr. L. Southerns .
Investigation as to [ Sept. 13 , calorimeter terminals were connected to these wires by tinfoil strips .
These precautions almost entirely did away with the disturbances .
The arrangements are shown in fig. 2 .
The counterpoise was also provided with an external vessel and gauge like that of the calorimeter .
With the modified apparatus a series of observations was taken , the pressure of air in the balance case ranging from 1-6 to 5'0 cm .
of mercury .
In one case 6 mm. was used , but a discharge took place within the balance case , fusing the tinfoil strips .
In this particular case the pressure was reduced by means of a bulb containing cocoanut charcoal , connected to the exhausting tube and cooled by liquid air .
The low vacuum , however , not sufficiently insulating the leads , the method was not repeated .
Final Experiments .
The table ( p. 399 ) gives a record of the final experiments made with the modified apparatus .
Observations were taken for a considerable period before and after heating ; the curves showing some of these are given in the diagrams indicated by letters in the last column of the table .
It should be noted that all the deflections marked in the table are temporary only ( except No. 8 ) and the balance righted itself in one or two minutes in all cases .
This effect plainly could not be due to an actual difference of weight due to alteration of temperature , or it would have persisted longer , for the fall of temperature in two minutes was very small\#151 ; always less than 1 ' C. Experiment No. 8 is abnormal , no doubt some accidental disturbance took place .
In calculating x in Column 9 the temperatures given in Column 7 have been used .
The calculation is as follows:\#151 ; Total weight heated = 300,000 milligrammes .
1 scale division corresponds to n milligrammes ( Column 8 ) ; .\ 1 " " to 1 in 300,000/ w ; .
' .
1/ 10 " " to 1 in 3,000,000/ w ; and if rise in temperature = 6 ' , 1/ 10 scale division corresponds to 1 in ( d x 3 x 106)/ ?i per degree rise of temperature , and as this is the smallest measurable deflection , we have x = ( 36/ n ) x 106 .
In nearly all cases a very slight oscillation of the beam took place during the passage of the current .
It seems reasonable to assume from the results of these experiments that any variation of weight which may occur is less than 1 part in 108 per degree rise of temperature .
The fact that all the small displacements observed were 1 .
2 .
3 .
4 .
5 .
6 .
7 .
8 .
9 .
10 .
11 .
12 .
13 .
14 .
No. of experiment .
Date .
Pressure in balance case .
Current ( alternating ) .
Time of passage of current .
Rise of temperature by platinum thermometer.* Rise of temperature by specific heat of the oil.* Sensitiveness of balance , 1 mm. of scale = n milligrammes .
Smallest measurable alteration of weight expressed as 1 part in x of total mass heated , per degree rise of temperature .
Alteration of weight during heating , in millimetres of scale , -f = increase , \#151 ; = decrease .
Alteration of weight during heating , expressed as 1 part in y , per degree rise of temperature .
Time , after cutting off current , taken by balance to resume its former position .
Remarks .
* The temperatures were taken by using the heating coil as a platinum thermometer , and by calculation from the known heating effect and specific heat of the oil .
The difference is probably due to non-uniformity of heating , as well as to an actual difference of temperature between the wire and the liquid in which it is immersed .
A in Column 10 indicates a quantity too small for measurement .
Refer to the curves marked with [ these letters .
i 1906 .
May 30 cms .
of Hg .
2*5 amps .
0-35 secs .
30 'c .
11 -8 'c .
9-0 n. 1 -o X. 2 -7 x 10 ?
mm. 0 iV .
mins .
No effect .
c 2 " 31 1 -6 0-35 30 11 -8 9 0 1 -o 2 -7 x 10 ?
0 \#151 ; \#151 ; 3 ' June 11 1 -6 0-35 30 11 -8 9 0 1 *0 2 -7 x 10 ?
\#151 ; A \#151 ; 1 Minute temporary deflection .
4 " 16 \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; Experiment spoiled by mercury in 5 " 22 1 9 0-35 30 11 -8 9-0 0-7 3 -8 x 10 ?
0 _ Gauge L , fig. 2 , going over .
No effect .
6 " 23 1 9 0-35 30 11 -8 9-0 0-7 3 -8 x 10 ?
-A \#151 ; 1 to 2 Minute temporary deflection .
7 July 23 3-2 0-35 30 11 8 9-0 0-86 3 -1 x 10 ?
0 \#151 ; \#151 ; No effect .
8 " 24 2 -4 0*35 45 \#151 ; 13 -5 0-86 4 -7 x 10 ?
-o-l -4*7 x 107 10 Bad result .
The only one which gave D 9 " 25 2 6 0-35 60 22 -5 18 -0 0-86 6 -3 x 10 ?
0 a permanent deflection with heating .
No effect .
E 10 Aug. 2 \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; Experiment spoiled as in No. 4 .
11 " 3 2-7 0-35 50 \#151 ; 15 -0 0-5 9 -0 x 10 ?
\#151 ; A \#151 ; 1 Minute temporary deflection .
F 12 " 4 2-6 0-35 60 22 -5 18 -0 0-5 1 -08 x10s -o-i -1 *08 x 10s 2 Small temporary deflection .
13 3 5 0-35 60 22 -5 18-0 0-49 1 -1 x 10s1 0 \#151 ; \#151 ; No effect .
14 " 14 4 6 0-15 360 \#151 ; 20 -0 0-6 1 -0 x 10s 0 \#151 ; \#151 ; j ) a 15 " 15 3-8 0-31 90 22 -8 21 1 0-6 1-05 x10s -0 1 \#151 ; 1 *05 x 108 1 Small temporary deflection .
H 16 " 16 3-9 0-27 140 26 -4 25 -2 0-6 l -26 x 10s 0 \#151 ; \#151 ; No effect .
J 17 \#187 ; 17 5-0 0-27 170 32 -5 30-6 0-72 1 -27 x 10s 0 \#151 ; \#151 ; \gt ; \gt ; K 1906 .
] Dependence of Gravity on Temperature .
399 Mr. L. Southerns .
Investigation as to 10-30 10-30 11-30 1906 .
] Dependence of Gravity on Temperature .
401 in one direction lias given rise to the suggestion on the part of Dr. Hicks that while no permanent effect is produced by rise of temperature , it is just possible that the body may be slightly lighter during the actual development of heat .
The present apparatus , however , is inadequate for the investigation of these minute effects ; some of them may be due to motions in the oil itself during the process of heating .
Comparison of Results with those of Pointing and A few words may be said with regard to the two methods employed .
The chief differences are , first , in the Birmingham experiments a much higher vacuum is employed ; secondly , the mass is of gunmetal , and is heated by means of an external steam jacket , and , thirdly , two experiments have to be performed , one with a solid mass and one with a hollow one , and the difference between the results taken , whereas in the present method only one is needed to give the complete result .
The advantages gained by the internal method of heating are , first , the actual experiment itself is completed in two or three minutes , whereas in the external method two experiments , lasting 8 to 48 hours each , are required , during which time the balance is swinging on its knife-edges .
This gives time for various disturbances due to external causes to interfere with the results of the observations .
Secondly , the external envelope of the calorimeter does not sensibly rise in temperature during the critical part of the experiment , and there is no necessity to use a high vacuum in the balance case .
This means that radiometer effects are not brought into play ( as is the case in the other method ) , whilst air currents are , nevertheless , practically or entirely absent .
Thirdly , the balance case , supports , beam , etc. , do not become heated , therefore there is no connection such as that which has to be applied on this account in the case of the external heating method .
In the experiments of Pointing and Phillips , the result is given as no variation of weight within 1 part in 109 per degree rise of temperature , while in the present paper a degree of accuracy of 1 in 108 only is recorded .
The Author , however , thinks that in point of delicacy the present experiments are probably at least equal to the above , on account of the fact that no corrections need to be applied to the results .
It would appear to be extremely difficult to apply satisfactory corrections for alterations of temperature of the balance case , and this has had to be attempted by them , on account of the considerable length of time occupied by an experiment .
On December 8 , 1904 , for example , the correction curve appeared to be 17 times as steep as on the day before .
Fig. 3 is plotted from the table on * 'Roy .
Soc. Proc. , ' A , vol. 76 , September 13 , 1905 .
402 Investigation as to Dependence of Gravity on Temperature .
p. 451 of the paper referred to and gives the correction for the hollow mass .
The observations made on particular days are connected by dotted lines .
The firm line shows the gradient deduced from the points marked by the method of least squares\#151 ; the application of this method , however , would appear to be superfluous in this case on account of the irregular disposition of the points .
The nature of the diagram will show how uncertain such a correction must be .
Also , the gradient is deduced from observations lying between 12a7 C. and 15 ' C. , but it is applied to all temperatures from 10o,45 C. to 170,45 C.* i.e. , over a range three times as great as that for which the observations for the correction were made .
At the higher and !
( fbal\amp ; rvc* .
Cclojl . .
Fig. 3 .
lower temperatures the corrections must be still more uncertain .
In some cases the correction deduced from a mean curve like that shown in fig. 3 ( it being impossible to obtain a fresh correction curve for each experiment ) amounted to over 60 times the final actual result ( 1 in 6 x 109 ) and in many cases to over 30 times this amount .
Again , it is impossible to measure directly the temperature of the gun-metal masses , and it appears probable that these would not assume the extreme temperature of the jacket , especially in the case in which cooling by liquid air was employed , when the time allowed for cooling was rather short .
* ' Koy .
Soc. Proc./ September 13,1905 , p. 453 .
There is evidently a misprint on line 3^of this page , which should read : Correction for temperature of case \#151 ; 0T3 division per 1 ? .
Continuous Rays in Spark Spectra of , etc. 403 The vacuum between the jacket and the mass would act as an efficient nonconductor of heat .
We may , however , probably conclude with considerable certainty on the combined testimony of the two series of experiments that , within the limits of temperature used , no variation of weight occurs greater than 1 part in 108 for a rise in temperature of 1 ' C. Note on the Continuous Rays observed in the Spark Spectra of Metalloids and some Metals .
By W. N. Hartley , D.Sc .
, F.R.S. ( Received October 17 , \#151 ; Read November 8 , 1906 .
) In a paper published in the ' Proceedings of the Royal Society , '* reasons were given for believing that the back-ground of continuous rays in the spark spectra of the metalloids , for instance , tellurium , arsenic , antimony , and bismuth , was caused by the light emitted by an incandescent oxide , whether in a state of vapour or solid , having its origin in the cooling of the dense vapour of the element in an atmosphere containing oxygen .
The spectra of metals which are not oxidisable did not show it , namely , gold , silver , and platinum , neither did those of the easily volatile metals such as mercury , indium , f thallium , zinc , and cadmium .
It was visible on photographed spectra of metals belonging to the iron group , but at the points of the electrodes only , where a non-volatile oxide is formed .
As the original explanation has not been accepted as satisfactory !
I have recently submitted the question to a special examination .
It must not be overlooked that spark spectra with continuous rays are yielded by the metals lead , tin , and cadmium , if the exposure of the photographic plate be increased to double that sufficing for the line spectra , and that sparks passing between electrodes of these metals in air deposit their oxides on all objects near to them ; in hydrogen , films of metal are deposited on the walls of the containing vessel , and when air is substituted for hydrogen there is at first a deposit of oxide , and subsequently one of * Yol .
49 , pp. 448\#151 ; 451 , 1891 .
t This word is misprinted t : iridium " in the ' Proceedings .
' X See Kayser 's ' Handbuch der Spectroscopie , ' vol. '2 , p. 286 ; also P. Lenard , ' Annalen der Physik , ' 1905 , vol. 17 , pp. 208\#151 ; 212 .
'

