]\gt ; THE effect of heat upon the modulus of torsional rigidity was first studied by in who investigated the changes in rigidity by observations on the torsional vibrations of wires .
This method is the one which has since found most favour with experimenters , being easier in manipulation and , perhaps , also capable of greater accuracy than the statical method of experimenting .
KUPFFER experimented on wires of iron , platinum , silver , copper , and gold , but the range of temperature employed was only that furnished by the varying temperature of the room\mdash ; never more than C. In a later paper an account is given of some experiments on wires of copper , steel , and brass at the temperature of the room and at that of boiling water .
The coefficient of diminution of the rigidity modulus was assumed to be constant between these temperatures , and was calculated from the periods of the torsional oscillations .
repeated KUPFFER'S experiments with wires of iron , brass , and silver : using a rarlge of perature of only a few degrees .
KOULRAIJSCH and LOOMISS experimentsd on wires of iron , copper , and brass at temperatures between C. and 10 C. , and came to the concIusion that for these metals the rate of alteration of the rigidity modulus with the temperature was not constant but increased with the temperature .
Their results showed that the rigidity modulus at C. could be represent fairly accurately in the , being the value of the modulus at C. obtained by exterpolation .
'Mem .
de l'Acad .
de St. Petersb VI .
ser. , , p. 233 , 1863 .
'Mem .
de l'Acad .
de St. Petersb VI .
ser. , .
VI .
, p. 397 , 1857 .
'Pogg .
Ann Ergsbd .
iii .
, p. 351 , 1853 .
S 'Pogg .
Ann 141 , p. 481 , or ' Amer .
Journ. Sci vol. 50 , 1870 .
V0L .
CCIV .
14.10.04 PISArI , a series of extremely thorough experiments , measured the torsional rigidity of various wires over a range of temperatures from C. to 30 C. He , too , found that the decrease of rigidity .
degree rise of temperature increased , at first fairly rapidly , with the temperature , but more slowly afterwards , the rate being practically constant above 19 C. The results of his experiments he expressed in a formula of the type TOMLINSON found that constant values of the rigidity modnlus could o1lly be obtained after the wire under test had been very carefully annealed .
He worked at four temperatures between and 10 C. , and his results are given in a formula similar to that adopted by KOHLRAUSCH and LOOMIS .
Experiments have also been made by KATZENELSOHN , GRAY , BLITHE , and DUNLOP , S and by SUTHERLAND the latter confining his researches to the softer metals , lead , inc , tin , and magnesium , which had been neglected by former experimenters .
worked at the ordinary temperature of the room , and at the temperature of boiling liquid air .
He used the statical method of experimenting by observations on the torsional deflections of a wire under a constant couple .
The experiments were extended to a large number of elements , but were taken at two temperatures only , and the rate of alteration of rigidity with temperature was assumed to be constant between them .
The main object of this research was to confirm the law that if the various metals are arranged in order of ascending temperature coefficients of the torsional modulus , they are also in order of increasing coefficients of expansion and of diminishing melting-points .
The gidity of copper and steel at C. has also been investigated by BENTON .
In view of the different results obtained by these observers , it appeared desirable to repeat the determination of the rigidity modulus at different temperatures , taking greater precautions to obtain a constant uniform temperature in the wire under test , timing the torsional oscillations with greater accuracy , and paying especial attention to the purity and physical condition of the wires used .
The experiments described in the following paper were performed on wires of various pure metals and some others , all of approximately the same diameter and length .
The method employed was a dynamical one , the torsional oscillations of the wire under test being timed by a method of coincidences capable of great exactness .
Some of the observations made 'Nuovo Cimento , ' ser. III .
, , p. 181 ; vol. 2 , p. 137 ; vol. 4 , p. 152 ; vol. 5 , pp. 34 , 136 , 145 ; 1877-79 .
' Roy .
Soc. Proc vol. , p. 343 , 1886 .
1 'Beibl .
zu den Ann XII .
, p. 307 , 1888 .
S ' Roy .
Soc. Proc vol. 67 , p. 180 , 1900 .
'Phil .
Mag [ 5 ] , vol. 32 , p. 31 , 1891 .
' Ann. .
Phys V. , p. 220 , 1901 ; IX .
, pp. 666 , 1124 , 1902 , ' Phys. Rev , p. 17 , 1903 , ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
in the course of this work yielded what seemed to be interesting information as to the internal viscosity of the wires used .
This is recorded and compared with similar observations by other imenters .
The rest of the paper is divided into the following sections:\mdash ; ( I. ) Description of apparatus , etc. ; ( II .
) Account of the experime1lts ; ( III .
) Summary of results and comparison with those of other observers ; ( IV .
) Delermination of the coefficients of expansion of vires .
PART I. DESCRIPTION 0F , ETC .
In these experiments the wire under test and the ibrator were both enclosed in a heating jacket , the temperature of which could be regulated as required .
This arrangement was the result of many experiments to find the best method of overcoming the various difficulties connected with obtaining a constant uniform temperature in the wire .
Some of these preliminary experiments will be referred to later , and the reasons for rejecting old forms of the apparatus will be given .
The part of the heating jacket surrounding the wire consists of two concentric tubes .
The outel one is a 3-inch brass tube , and is covered extel.nally with layers of cotton wool to a thickness of an inch all round .
The inner tube is of copper inch in thickness and 1 inch in internal diameter .
This is joined to two short lengths of wider copper tube at the top and bottom , in order to admit the clamps for the wire .
The inner tube of the jacket projects about inches beyond the outer one at the top , and is there held firmly in a vertical position by a wrought-iron clamp , I , fixed to the wall of the laboratory , and projecting about 6 inches from it .
The heatel is also supported by a larger clamp which grips the outer cylindrical tube just above the box containing the vibrator .
The vibrating plate hangs inside the double-walled copper box F. The floor of this , requiring to be rigid , is made of -inch metal , but the other walls are not so thick .
The edges were , unfortunately , soldered together instead of being brazed , and this caused many weeks of delay , the apparatus having to be epeatedly resoldered to prevent leaking at the higher temperatures , due to unequal expansions tearing some of thejoints asunder .
As will be seen from the figure , the double walls of this box are in direct communication with those of the cylindrical part of the heater , so that the steam by which the jacket is heated flows through the two .
It is , of course , necessary to have a door to the front of the box .
This also has double walls and is fed with steam by means of a side tube from the cylindrical part of the heater .
The waste-pipe from the door joins that from the bottom of the box , as is shown in the figure .
These pipes are made of -inch ' ' composition\ldquo ; and are joined on to the door by means of brass " " unions a thin washer made of sheet asbestos being used to make the joint tight .
The door and the front of the box on which it fits have ON THE NODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
to the axis of the rod , which was then slit down to this point with a fine saw .
very thin piece of metal was then inserted in the slit , the collar placed on and screwed up tightly by he steel screw , and a hole bored down the axis of the rod of the same size as the wire to be used ( No. 20 , B.W.G. ) .
In this way a neat clamp is formed which grips the wire all around its circumference at a perfectly definite point , viz. , the ends of The lower end of the wire is held in a clamp exactly to that at the top .
This clamping rod has a screw thread at its lowel end by which the vibrator is fixed to it , being screwed home against a projecting flange .
Below the screw thread , the rod was turned down to inch in diameter and one half cut away , so as to leave a plane face for affixing the mirror used in timing the vibrations .
This mirror was held in a light brass frame which could be fixed in a perfectly definite position on the clamping rod by means of two small screws .
The vibrating plate hangs about 1 inch below the top of the box , and above it a ring of ' gunmetal is supported on two horizontal glass rods going from side to side .
This ring is for determining the moment of inertia of the vibrator , and can easily be lowered by hand Fig. 2 .
on to the plate .
The plate is a circular disc of gunmeta ] inch thick and 5 inches in diameter .
The outer edge of the upper face is turned down about inch .
The ring exactly fits on to the groove so formed , and there is thus no difficulty in placing it centrally on the plate .
In order to set the plate in vibration without swinging it like a pendululn bob , the following device was adopted .
Two glass jets inclined at 45 degrees to the vertical in opposite directions are fixed underneath the plate , and about inch below it , one at each end of a diameter .
These two jets are connected together at the bottom of the box by a glass tube which passes out through the side , and is joined outside to 3 or 4 yards of fine " " composition\ldquo ; tubing , which passes along the walls of the laboratory and ends on a small table beside the observing telescope .
Here it is joined by means of a piece of indiarubber .
to a small glass funnel , and by depressing this in a dish of mercury a wind is sent through the jets which puts a couple on the plate and rotates it .
By properly timing the instants at which the funmel is loweled into the mercury ( by watching through the telesco ] ) ) , the amplitude of vibration of the plate is quickly increased .
The Observations of the periods of vibration of the wires tested were made in general at five temperatures , viz. , at the temperature of the room , about , at C. , C. , C. , and 10 C. , and also in some cases at hese t } DR. F. HORTON ON THE EFFECTS OF CHANGES OF TEMPERATURE were indicated by two delicate mercury thermometers , one of which had its bulb in the cylindrical copper tube close to the wire and at the middle of its length ; and the other protruded horizontally into the copper box containing the vibrator , and passed a few millimetres below the plate , the bulb being at the middle of the box ( see and , fig. 1 ) .
These thermometers were graduated to C. , and readings to C. could readily be made .
In order that the readings of the thermometers might not be affected by the varying temperature of the room , the parts of the stems protruding from the jacket were enclosed in glass tubes about an inch in diameter , through which a constant stream of water was kept flowing .
The temperature of this stream was ascertained by means of separate thermometers , one in each waterjacket .
In the observations at laboratory temperatures the heating jacket was connected to a water tap , and a rapid stream of water allowed to flow through it .
It would then , in general take three or four hours for the temperatures , as indicated on the thermometers , to become constant .
The upper one , as would be expected , arrived at a steady state long before the lower , but in no case were observations taken until both had attained a constant temperature .
The necessity for this precaution is at once seen , for if the temperature of the thermometer is varying , then it is most probably different from that of the plate or wire , as the case may be .
Also , if the plate is at a different temperature from the wire , there must be a flow of heat taking place along the wire , which consequently cannot be at a uniform temperature .
The other temperatures were all obtained by using the vapours of various liquids boiling under atmospheric pressure .
The liquids used were ether , acetone , methylated spirit , water , and amyl alcohol .
In order that the composition of the liquid , and therefore its boiling-point , might not vary , a reflux arrangement was used .
This is shown in fig. 1 .
The vapour is driven from the large copper boiler , through the tube , into the top of the heating jacket .
It passes out at the bottom , and is condensed in the water condenser , the condensed liquid running back into the boiler .
The boiler is open to the atmosphere through the tube , which is also surrounded by a water-jacket , and serves as a safety tube .
The tube is a 1-inch " " composition\ldquo ; tube , and is fitted to the heating jacket by means of a brass union joint , a washer made of thin sheet asbestos being used to make the joint tight .
The tube was well wrapped up in cotton wool , this being covered externally with a cylinder of asbestos cardboard\mdash ; as was also the whole of the cotton.wool round the heating jacket\mdash ; in order to prevent it from taking fire in case of an accident .
In the first form of the apparatus , the wire alone was surrounded by the heating jacket , and the vibrator hung inside a wooden box , the temperature of which was irxdicated by a thermometer passing close to the plate .
Happenin one day to put the thermometer at a further distance from the plate than usual , I was struck by the difference in the temperature indicated .
This led me to experiment with several ON THE MODULUS OF TORSIONAL RIGIDITy OF METAL WIRES .
thermometers at different distances from the plate , and the results showed that the plate was much hotter than the rest of the enclosure .
The seriousness of this is at once evident , for not only were the temperatures which had been taken to be those of the vibrator too low , but heat must have been flowing from the wire , which , therefore , could not have been at the temperature of the jacket enclosing it .
It was first attempted to stop this leakage of heat from the wire by interposing a rod of some non-conducting material between the vibrator and the wire .
Some experiments were made on the conductivities of various materials , and finally slate was selected as bein the best from a workable and non-conducting point of view .
A rod of slate , 4 inches long and inch diameter , was fitted between the vibrator plate and the clamp which held the wire .
Even with this arrangement the plate warmed up a little , about C. , when the heatingjacket was at .
C. In order to see how this affected the distribution of temperature along the wire , a copper wire was suspended in the heater , and to this were soldered three nickel wires at different points along its length , thus giving three pairs of thermo-electric junctions , by means of which any temperature differences could be detected .
The results were surprising .
A poin inch from the lower clamp and 4 inches inside the heatin iacket , was found to be at a temperature C. below that of the centre of the wire , while at a point on the wire , 6 inches inside the heater , this difference was C. These experiments indicate that the only way to obtain a uniform temperature in the wire is by enclosing the whole vibratin system in the heatingjacket .
The interposition of non-conducting material reduces the temperature differences , but does not entirely do away with them .
As the heat had been leakin out of the wire through the bottom clamp , it must also have been doing the same through the top one .
The top clamp was therefore severed from the bar ( fig. 1 ) , and a piece of slate , 8 inches in length , interposed between the two .
The rod was then lowered further inside the heater , and the difference of temperature between the upper end of the wire and the centre was determined .
No deflection could be obtained on the galvanometer used , showing that the temperature difference was less than C. After the new heater described in this paper had been made , the distribution of temperature along the wire was again tested , but no difference could be found between that at the centre and that at either end .
The piece of slate interposed in the upper clamping rod unfortunately cracked with the heat after being used for a few weeks , and a similar one of marble was therefore substituted .
This was exceedingly difficult to turn owing to its brittleness , but it has successfully withstood the effects of temperature for over a year , while further experiments with the thermal junctions proved it to be as effective as slate in preventing a leakage of heat .
It might here be mentioned , that this research was commenced at the Physical Laboratory of the University of Birmingham .
In the form of apparatus used , the wire experimented on was 110 centims. long .
On starting afresh at the Cavendish Laboratory\mdash ; where most of the work described in this paper was carried out\mdash ; this size was reduced to 58 centims. , the smaller heating jacket thereby required )eing more convenient , in that it comes to a steady state more quickly and is more easily maintained at a constant temperature .
Correctioms to the Thermometer The thermometers used were four in number , two ranging from to 10 C. , and two from to C. , all graduated to .
These were standardised at the National Physical Laboratory , the tests being carried out with the thermometers under exactly the same conditions as obtained in these experiments , viz. , in a horizontal position , and with nine divisions in the hot bath , and the rest of the stem surrounded by a water jacket at a known temperature .
From the figures supplied by the National Physical Laboratory , curves of corrections were drawn , from which the corrections to the thermometer readings at any temperature could be read off with an accuracy of C. A small correction had to be applied on account of the difference in the temperature of the water-jacket in these experiments and when the thermometer was standardised .
This was found by calculation , taking the coefficient of apparent expansion of mercury in glass as 00016 .
The correction on this account was never greater than C. The method of heating with the vapours of boiling liquids worked extremely well .
The jacket would take from 3 to 4 hours to arrive at a constant temperature , and then neither thermometer would vary by C. even in the worst cases , and it often happened that both thermometers remained absolutely steady for hours .
It is a curious fact that the corrected readings of the two thermometers weoe rarely the same .
The difference was generally less than C. , but sometimes the upper and sometimes the lower one would indicate the higher value .
In calculating the results , the temperature of the wire was always assumed to be that of the upper thermometer , while the lower thermometer was taken as indicating the temperature of the vibrator .
the by Method of Coincidences .
The method used consists in observing the reflections of a vertical llash occurring once a second , in two mirrors , one of which is fixed in position , and the other is attached to the vibrating plate and vibrates with it , swinging just above the fixed mirror and being parallel to it when at rest .
The reflections are observed by a telescope , in the field of view of which , in general , two flashes are seen , one always occurring in the same position , and the other appearing in different parts of the field , according to the position of the moving mirror at the instant the flash occurs .
If a ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
second signal happens exactly when the two mirrors are parallel , the two flashes coincide , and it is from these " " coincidences\ldquo ; that the time of vibration is obtained .
The method of coincidences is usually only applied to the comparison of two nearly equal times , but Professor pointing suggested to the author a modification of the method which can be applied to two periods , even if they are quite different .
In order to illustrate this , let us assume that the time of a complete vibration is .
seconds .
This has to be compared with the 1 second period of the flashes .
Now suppose that the two images of the flash as seen in the telescope have just coincided , and let us call the second at which this coincidence occurred .
Then after 4 seconds the two mirrors will not be exactly parallel again , and the flashes observed in the telescope will consequently be some distance apart .
Since the period is greater than 4 seconds , the moving mirror will not yet have become parallel to the fixed one , and the moving flash will appear to have fallen short of its zero position .
If , however , we wait for such a number of seconds , , as is very nearly an exact multiple of the time of vibration , then at the flash the two mirl.ors will be very nearly parallel , and the flashes will very nearly coincide .
No ) ; therefore after 37 seconds , the moving flash will appear very near indeed to the fixed one , for it would take the swinging only 044 second to become parallel to the fixed mirror .
As before , since the multiple of the time of swing is ) reater than 37 seconds , the flash will appear to have fallen short of its position of rest .
Now suppose we on counting the seconds , calling the flashes next after the 37th , one , two , three , \amp ; c. , up to 37 and then st , arting at one again , and so on .
Every 37th flash will appear to have lost on the position of the preceding.37th , , to have moved further away from the central fixed flash in the opposite direction to that in which it would appear to be moving if the light were continuous .
As this goes on flash number 1 next after will be nearer and nearer to the central fixed flasl ] , until after a time it coincides with it .
Suppose that when this happens counted sets of.37 econds .
Then it is evident the mirror must have lagged behind one complete second , for it takes one }econ d longer for it to become parallel to the fixed mirror .
The vibrator therefore makes )rations i seconds , and if is the period of vibration we have , or In general an exact coincidence is of very rare occurrence , and the nearest coincidences are taken .
Thus suppose A is the position of A the fixed flash as seen in the telescope and that a 37th flash appears at , and the next 37th at C. Then the mirrors must have been parallel at the fraction ( say ) of a period after , so that the coincidence period is instead VOL. CCIV.\mdash ; A. of , and we have .
In order to measure AB and BC the telescope used had a fine horizontal scale in the eyepiece , and , as a general illumination of the field of view of the telescope made it difficult to see the flashes , the scale was illuminated by the light from a small ectangular slit in a dark screen placed in front of a lamp .
The light was reflected into the telescope by means of a small right-angle prism , a piece of ground glass was placed in front of the slit to make the illumination more uniform , and also a piece of reen glass , for it was found that the flashes could be best seen when the scale was illuminated with this colour .
In order to see what degree of accuracy this method is capable of , it must be noticed that an error can only arise in the fraction only an inexcusably careless observer could be 1 out in the value found If we take as 20 ( a mean value ) we see that approximately ; the occurs in and is therefore not greater than secoml .
If is less than 20 it follows that the flashes are " " losing\ldquo ; more rapidly , and hence the distances to be measured in the telescope are larger , and the value of can be more accurately found .
If on the other hand is larger than 20 , the distances between two consecutive 37th flashes is smaller , and consequently is more difficult to measure ; but in none of the cases dealt with could the have been as large as 1 , and it must be obsel v the larger becomes of less importance is the fraction In what has been said above it is assumed that the vibrator is gradually losing on the 37 second period .
This is of course not always the case .
It may be gaining and making 9 ations in a little undel 37 secconds and 9 vibrations in 37 seconds ; whence In this case each 37th flash appears to gain on the previous 37th , and so to move out from the centre of the field in the opposite direction to that already considered .
After periods it is now the ) flash which nearly coincides with the fixed one .
coincidence period is of course not always 37 .
To find it the time of vibration must first be approximately kuown .
This is found by GAUSS'S method , using a chronometer .
Suppose it to be 4 1154 seconds .
Such a number of seconds is required as is ON THE MODULUS OF TORSIONAL IDITY OF WIRES .
nearly an exact multiple of time .
Considering the fractional part only , put , whence successive convel.gen ts to which are Hence the numbers which multiplied by products neal.est whole numbers are 8 , 9 , 17 , 26 . .
The larger the factor the nearer is the product to a whole number , and at first sight it might seem to ) of advantage to use the largest to obtain the coincidence period , but taking 26 we have , so that in a period of 107 seconds the vibrator will have " " lost\ldquo ; 0004 second , and therefore to lose 1 second it will take 2500 periods of 107 seconds , or over 70 hours !
In the present instance 9 is the most suitable factor and gives , so that in a period of 37 seconds the vibrator lags behind by 0386 second , and therefore ' ' loses\ldquo ; 1 second in 16 minutes .
The The second signals are given by a clock pendulum , carrying at its lower end a platinum wire , which cuts through a fine stream of mercury each time the pendulum passes its position of rest .
This " " makes\ldquo ; circuit , which by means of a relay " " breaks\ldquo ; another circuit through the prinlary of an induction coil , thus giving a flash in a vacuum tube connected to the secondary circuit .
The platinum wire is flattened out perpendicularly to the plane of swing of the pe1ldulunl .
It is 6 millims. long and is screwed into a small brass pin , which fits into hole in the lower end of the pendulum , being secured ) osition by a ) screw .
lt call easily be emoved f cleaning .
This pin isjoined to a fine coppel wire running up the back of the wooden pendulum rod and the top to the lo end of the steel strip which the pendulum is suspended .
The ) suspension is connected by a to a terminal screw outside the clock case .
The stream of mercury issues from a very fine glass jet fixed to the back of the clock case .
An adjustment in the position of this jet call be made by means of two large milled-headed screws , so as to direct the stream exactly through the position of the platinum wire when the pendulum is at rest .
The jet is at the end of a glass tube which enters the case through a hole in the side , and is bent so as to direct the stream of mercury perpendicularly to the plane of swing of the pendulum .
This tube is connected by indiarubber to a reservoir containing some 500 ) .
centims. of mercury , at a level of about 30 centims. above the jet .
The rate of flow of the mercury can be regulated by a stop-cock .
In the glass tube , just before the jet , are two air , in order to stop any bubbles of air which might interrupt the mercury stream .
A platinum wire , connected outside to a terminal , is also sealed into the tube to make contact with the mercury .
The mercury running from the jet is collected in a wooden box from which it can be drawn off and returned to the reservoir .
The clock break is connected in series with ] cells and a contact breaker , shown in fig. 3 .
The wires from the clock are connected to the terminals of the small electromagnet A. Every second this becomes excited and attracts a soft iron armature , , which is fixed to the shorter arm of the light brass lever C. This lever is 18 centims. long and the short arm is centims. To the longer arm two platinum wires are brazed , one at and the other at E. The wire at dips into a mercury cup , , and is not drawn out when the magnet is excited .
The wire at dips into another cup of mercury , , and the position of the lever is so adjusted , by the two stops and , that is drawn well out of the mercury in when the clock circuit is made .
The terminals and are connected to the mercury in the cups and respectively .
Wires from them go to the primary of the induction coil , two accumulator cells being included in this circuit .
The contact breaker on the induction coil is tied back so that it cannot work , and hence once a second the primary circuit is broken and a flash occurs in the vacuum tube connected to the secondary .
At first a layer of water was kept over the surface of the mercury in , but after a time this became very dirty ( the " " dirt\ldquo ; being probably a platinum amalgam in a very fine state of division ) and a skin formed on the surface of the mercury .
This led to the circuit sparking two or three times at each break , and caused much trouble , for the result is two or more flashes in the vacuum tube ( not separately visible ) , and therefore two or more in different parts of the field of view of the telescope .
In order to overcome this difficulty , the arrangement shown in the figure was used .
The cup was made by blowing a hole in the side of a bulb blown in a glass tube , one end of which was then bent up ( P ) and the other slightly down ( Q ) .
The mercury in does not come up to the level of these side tubes .
is connected to the stop-cock of a vessel containing water and hangs oyer a bucket .
The water flows slowly into and its level gradually rises until the extra pressure sufficient to overcome the surface tension of the drop which always hangs from Q. It then ON THE MODULUS OF TORSIONAL IDITY OF 1.3 runs out with a rush , leaving only thin layel ' water above the the mercury .
This is repeated , and the ternittent 1 of water is quite efiectu in keeping the surface of the mercury clean .
of the lever on hitting against the stops was another difficulty which caused double sparking .
This was overcome by covering the stops with cork .
In order that the platinum wire at the end of the pendulum should make good contact with the mercury , it was occasionally removed , scraped , heated , and amalgamated with sodium amalgam .
It was then gently heated again to UIlamalgamate it .
This method\mdash ; recommended by C. .
Boys*\mdash ; was found to give the best results .
The wire the contact breaker had to be periodically taken out and ground smooth on an oil stone , for the sparking made the end ragged .
It was then heated and treated in the manner described above .
In order to prevent sparking at the clock break , a battery of 10 small cells ( connected up in sets of two in series and the sets joined up in parallel ) , consisting of strips of platinum foil in accumulator acid , were placed in parallel across it .
These soon become polarised , and their back E.M.F. prevents the current from the accumulators from going through them , but the extra E.M.F. when the clock circuit is broken sends a momentary current through these cells instead of sparking across the air gap .
Several other methods of obtaining second signals were tried with different kinds of clock contact and various forms of relay , but after giving much trouble they were discarded for the present arrangement , which has proved very satisfactory .
I had expected that the stream of mercury would affect the rate of the clock , but on rating it for several days with the mercury continuously from 10 .
till 10 , no difference in the rate could be detected .
The clock rate was determined daily by comparing the time with that indicated by the clock in the Cambridge Philosophical Society 's Library .
This latter clock keeps very good time , and is checked once a week by a man from the observatory .
Its weekly rate is about 2 seconds .
Twice a day\mdash ; at 10 .
and 6 .
\mdash ; a chronometer was carried across to the library and its time compared with that of the Philosophical clock , and then immediately afterwards with the clock in the Cavendish Laboratory .
The rates thus obtained were afterwards corrected on account of the rate of the library clock .
The clock used in these experiments was found to keep a constant rate for a week , altel.ing slightly each time it was wound up .
The vacuum tube giving the second signals was fixed vertically at a distance of about 4 metres from the mirrors .
The light from it was focussed by means of a glass cylinder full of water on to a slit made by cutting a fine line on the back of a silvered mirror .
The images of the slit in the mirrors were viewed by a telescope placed 80 centims. from the mirrors , the slit and telescope being so arranged that the incident and reflected light were nearly normal to the mirrors .
' Phil. Trans , vol. 186 , p. 26 .
Ths hxed nlirrol was supported on a small brass stand esting by.three screws on the ' of the ting iacket ( see fig. 1 ) .
It could be rotated about a in order tu set the )rallel to the one on In general , on one day , two sets of observations were taken , the first at the temperature of the tap water , and the second at some higher temperature .
In order to get the two done in one day it was necessary to leave the water running through the jacket over night , so that in the morning no time should be wasted in waiting for the temperature to become steady .
A set of the teml ) el.ature of the tap water having been taken , the heating jacket was emptied and dried by sucking warm dry air through it .
It was then connected up to the boiler in the manner already described .
As a rule , it would take from 3 to 4 hours for the temperature to become constant , and in the interval the rest of the apparatus was tested and set in order if necessary .
The fixed mirror had to be re-set parallel to the moving one at each telnperature , for the wire altered its zero position slightly .
When the temperature had become constant the plate was set vibrating , the angle through which it moved being observed by placing a candle behind the slit so as to illuminate it continuously .
The amplitude of oscillation of late eased until the imag .
the slit moved ovel slightly more than divisions of the telescope scale .
The ] then left ) for a short time in order to the .
effects of .
The amplitude of swing having danll ) down to 22 divisions on the eye-piece , the time of a chronometel ' noticed , and the temperatures of the thermolneters read .
This was a determiuation of the logarithmic decrement of the oscillations which was made with each observation of the period .
When the amplitude had further damped down to 20 divisions of arc the candle emoved f behind the slit the observations of the period begun .
Since it was times of complete vibrations that were required , a coincidence being taken when the moving mirror was parallel to the fixed one , nd moving always in the same direction , an error might have arisen through variations in the zero position of the wire during an experiment .
In order to eliminate this , two sets of coincidence periods were taken in which the flashes either " " lost\ldquo ; or " " gained\ldquo ; in opposite directions .
Everything being ready , and the temperature of the thermometers noted , the observer sits down to the telescope and watches until he sees a coincidence is about to happen .
He notices one flash is " " early\ldquo ; by a certain amount while the 37th ( if 37 is the coincidence period ) is " " late\ldquo ; by another amount .
These are noted down He must then on counting from 1 up to 37 continuously , making a note of each 37th .
After a time he will observe that a coincidence in the opposite direction is about to happen , i. e. , that some flash\mdash ; say the \mdash ; is moving across the field in the opposite direction to the 37th and will soon be coinciding with the fixed flash .
This ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
coincidence is noted in the same way as the former one .
Then , if the flashes are " " losing after a time the 37th flash will have lost nearly a second and flash number 1 will be ut to coincide .
This coincidence is waited for and noted .
In the same way later on the gets a second behind and the 24th coincides .
It is only necessary to keep watching the flashes just when a coincidence is expected , and in the intervals between the coincidences the thermometers are read every few minutes .
In order to make sure that it has not been omitted to make a note of any 37th flash , the time of the chronometer at which one of these flashes happens near the beginning of the observation is noticed , and also again at the end .
From the total time the number can be checked .
After the observation of the period was finished , the candle was again placed behind the slit and the amplitude of vibration again observed .
When this became an exact number of divisions of the eye-piece scale the time of the chronometer was noticed .
From the times at the beginning and the end , and the calculated period of vibration , the number of vibrations could be obtained , and heuce the logarithmic decrement of the swings calculated .
The of Damping on the Period of Vibration .
I had anticipated having to correct the times of vibration on account of the damping of the amplitude , but in nearly all cases the too small to take into account .
If the equation of motion of the disc is the period of damped vibration is given by ' where is the logarthmic decrement of the amplitude of vibration .
Now if I is the moment of inertia of the vibrating system , and if is the rigidity modulus , the length , and the radius of the wire , we have Hence ; so that 8 Il If now we correct , the observed period , to , the time of vibration the system would have if there were no damping , quating these two values of 7 ?
, DR. F. HORTON ON THE EFFECTS OF CHANGES OF TEMPERATURE Except in the case of aluminium and of the soft metals tin , lead , and cadmium , the value of was never greater than , and it was generally much less than this .
Taking 15\mdash ; the value for a gold wire at C.\mdash ; the period was : seconds so that or the correction in this case is negligible .
In the case of the soft metals it was found to be impossible to time the vibrations to such a degree of accuracy as was obtained with the other wires , and only in a few ces was it necessary to apply the correction for damping .
Th.e Moments of of the The same vibratol was used for observations on wires of copper , , steel , silver , platinum , gold , and aluminium .
It consisted , as already described , of a circular plate of gunmetal screwed through its centre on to the lower end of a steel rod , the upper end of which was formed into a clamp to hold the wire .
The total mass of this vibrator was about 1 kilogralnme .
Lighter vibrators had to be used in the case of the softer metals .
With cadmium a rectangular sheet of brass weighing 100 grammes was substituted for the circular gunmetal disc .
This was replaced by a similar sheet of aluminium in the case of tin and of .
The moments of of these two latter vibrators were found with sufficient accuracy by calculation from their masses and dimensions ; } ) more exact knowledge was required in the case of the circular vibrator on account of the greater precision with which the torsional periods of the harder wires could be determined .
The moment of inertia of the gunmetal vibratol was therefore deterlnined experimentally , the on which it was suspended being a steel one .
Attempts were also made to find the moment of inertia of the vibrating system in the case of other wires , but it was found to be impossible to do so , for in nearly all the wires ated the rigidity was not a constant quantity , but varied very mnch even one teln pernture , so that tions made at differe1lt times could not be collpared .
Also in the case of the metals copper , gold , and platinum the extra weight of the ring used in determining the moment of inertia altered the rigidities of the wires very considerably , so that the periods with the ring on could not be compared with those of the vibrator alone on this account .
The experiment was tried with each wire , and it was found that the time of vibration of the plate alone , before the ring was put on , was largely different from that after it had been taken off .
The figures obtained in the determination of the moment of inertia of this vibrator , and its variation with the temperature , will be given later .
Lengths of the Wir The wires experimented on were all about 58 centims. long .
For the purpose of measuring .
lengths , a fine mark had been ibed on the upper clampin rod ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
( fig. 1 ) , and also another on the lower rod , beneath the gunmetal plate and just above the mirror .
This latter mark could be seen through the glass window in the door of the heating jacket .
The distance between these two marks was read every few days by means of a cathetometer .
The lengths from the marks to the jaws on each of the rods and had previously been measured , and by subtracting these distances , corrected for any difference of temperature , from the total length , the length of the wire at the temperature of the observation was obtained .
The radii of the wires were found by weighing a known length of the wire first in air and then in water .
The ends of the wire were filed square with the length , which was then accurately measured .
The wire was then wound up in the form of a spiral and its volume found by weighing in air and afterwards in distilled water at a known temperature .
The lengths of the clamping rods , of the wires in the determination of their radii , and the dimensions of the ring used in finding the moment of inertia of the vibrator , were all measured on Professor POYNTING'S measuring bench in the Physical Laboratory of the Birmingham University .
This consists of a trolley running on a railway underneath two vertical microscopes , which can be placed at any desired distance apart .
The microscopes have each a rocking-plate micrometer , by means of which a length of 001 millim. can be measured .
The article whose length is required is placed under one microscope and a standard metre under the other .
The trolley is moved until the desired point on the article is under the first microscope , and the reading of the standard metre under the second is noted .
The trolley is then pushed until the other point comes under the first microscope , and the scale is again read through the second .
In this way the distance between the two points is accurately determined .
The length of the standard metre was corrected to , at which temperature it is certified correct .
In obtaining the dimensions of the moment of inertia ring , two diameters at right angles were measured , and then the ring was turned over and the same two diameters measured from the other side .
The weight of the ring was grammes .
The mean radii were centims. and centims. at C. Hence the moment of inertia at C. .
cm.2 Corrections to the Observed Periods of Vibration .
In order that the observed periods at different temperatures may be comparable , it is necessary that they should be corrected for the increased length and radius of the wire , and also for the expansion of the vibrator , at the higher temperatures .
Two other corrections might have been uecessa1y , one for the logarithmic decrement of the * For description , see POINTING , ' Phil. Trans , vol. 182 , p. 589 , 1891 , or ' The Mean Density of the Earth p. 96 .
VOL. CCIV.\mdash ; A. oscillations , and the other for their amplitude .
The former was generally too small to take into account , and the latter was avoided by always dealing with the same amplitude .
The formula connecting the period of vibration with the rigidity and dimensions of the wire , and the moment of inertia of the vibrator is , where is the moment of inertia of the vibrator and the modulus of rigidity , being the radius and the length of the wire .
If now is the coefficient of expansion of the material of the wire , and that of the material of the vibrator , and if the wire is at , and the vibrator at , we have , where is the rigidity modulus at .
Hence .
In order , therefore , to correct the observed periods to the values they would have if the wire and vibrator retained the dimensions they had at , they must be multiplied by .
In the experiments C. was taken as the temperature of comparison , and the observed times were corrected by multiplying by The value of and the various values of for the different wires were determined experimentally .
The vibrator , as already stated , consisted of a gun-metal plate with a steel rod through its centre , and the temperature coefficient of the moment of inertia of this system was required .
This was found in the following manner , a steel wire , the temperature coefficient of the rigidity modulus of which had already been found , being used to suspend the vibrator : vibrations were timed at the temperature and then the ring was lowered on to the vibrator and the period again determined .
The ring was next turned round about a vertical axis through an angle of and the period found again , then turned through a further , and so on , taking six observations in all , the mean of which was taken as the period of vibration of the vibrator ring .
The ring was lastly taken off the plate and the period of the vibrator again found , the mean of this and the first determination being taken as the period of the vibrator alone .
Each time the heating jacket was opened to move the ring it had to be left for from one to two hours for the temperature to become constant again .
On the following day the temperature of the jacket was raised to , and two sets of observations with the ring on and off were taken .
It had been intended to take more , but the personal discomfort involved in removing the ring from the plate whilst hot persuaded me to be content with two .
The coefficient of linear expansion of the material of the ring was assumed to be the same as that of ON IODUL U OF TOliSIONAL RIGIDITy OF METAL WIRES .
a long gun-metal bar , which had been cast at the same time as the ring .
The coefficient of expansion of this bar , as well as the coefficients of expansion of the various wires , was determined in the manner described in Part IV .
of this paper .
The results of the above observations of periods are given below:\mdash ; Mean of the periods with the ring on seconds at C. ; C. Mean of the periods with the ring seconds at C. ; C. The mean period with the rin on must now be corrected to the value it would have if the temperatures had been the same as when the ring was off .
Correcting therefore on account of the alteration of rigidity of the wire , and of the dimensions of the wire and plate , we get the corrected time seconds .
Calling the period of the vibrator alone , and I the moment of inertia of the ring , we have or The value of I at C. was calculated from the known value at C. the experimentally determined value of the coefficient of expansion .
Hence at C. .
cm.2 The values of at the higher temperature were calculated in an exactly similar manner , and the results obtained were From the first experiment , c at C. From the second experiment , grm cm.2 at C. From each of these values , and that given above for C. , the coefficient of change of moment of inertia of the vibrator with temperature was found .
The values were00003540 and 00003501 ; mean ; from which required in the formula for correcting the periods is 00001760 .
PART II .
0F THE EXPERIMENTS .
The wires experimented on were , except in a few cases , chemically pure , and were supplied by Messrs. JOHNSON , MATTHEy \amp ; Co. , of Hatton Garden , London .
They were all of approxinuately the same length and diameter .
In order to get them into comparable conditions , they were cal.efully annealed before the rigidity determinations were begun .
This was done by repeatedly heating with an electric current , the oxidisable wires , iron and steel , being enclosed in an atmosphere of pure dry hydrogen , copper in carbonic acid gas , and the others in air .
In each case about a metre of the wire was suspended vertically in a glass tube , a light weight being attached to its lower end .
This closed an electric circuit by dipping into some mercury at the bottom of the tube .
In the cases of steel and of iron the tube was then repeatedly exhausted and filled with pure dry hydrogen until all the air had been removed .
Carbonic acid gas was used instead of hydrogen in the case of copper ; gold , silver , platinum and aluminium were heated in air .
The wire was then repeatedly heated to a red heat by means of an electric current , each heating lasting for several minutes .
The temperature attained was of course highest in the case of those metals which are furthest from their melting points at ordinary temperatures .
Platinum was taken almost to white heat , and iron and steel well above the temperature of recalescence .
After this treatment , a length of about 60 centims. was cut from the middle of the annealed wire and suspended in the heating jacket .
All the specimens used were quite straight and free from kinks , and special care was taken to see that the wire had not been heated to too high a temperature , or subjected to any permanent elongation by the tension to which it was subjected during the annealing .
The wires were all bright and free from oxide .
The softer , lead , tin , and cadmium , were annealed by suspending them ( without weight attached to their lower ends ) in a vertical combustion tube , heated by means of vertical rows of small Bunsen burners .
A thermometer was suspended beside the wire , and the temperature was not allowed to exceed two-thirds of the melting point of the metal in question .
The temperature was kept up for about an hour , and then the tube was allowed to cool down very gradually .
The heating and cooling were repeated many times .
Some unannealed wires were also experimented on .
Before giving the results in detail for the different wires , it would be well to mention that in case examined , with the exceptions of pure copper and steel , it was found that the modulus of rigidity at a constant temperature was not constant , but increased as time went on .
This increase was generally very small , and was greater at the higher temperatures than at the ordinary temperature of the ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL laboratory .
After repeated heatings for long periods to one temperature , the increase with time at that temperature gradually became smaller and smaller , and in some cases became zero ; but if then the wire was heated to a higher temperature and subsequently brought back to the lower one again , the increase with time at that temperature still went on .
This effect could not be due to any disturbing cause , such.as the oxidation of the surface of the wire or a gradual lengthening under the tension , for these would have the opposite effect , and cause an apparent diminution in the rigidity .
The wires were quite bright at the end of the experiments , and the length was mea , sured at intervals and in many cases was quite constant .
In the case of gold and the softer metals , however , it increased slightly during the experiments , and the periods obtained were corrected on this account .
Under these circumstances it is evidently impossible to mine accurately the manner in which the rigidity is affected by the change of temperature , for such change is complicated by the varying time effect .
It was thought that the most useful information would be obtained by plottin the values of the rigidity modulus against the corresponding , and drawing a curve through the first points obtained at each temperature .
The alteration of rigidity per 1o C. rise of temperature at the ordinary laboratory temperature was then found by drawing the tangent to this curve at C. In this manner a coefficient was obtained each wire such that for temperatures in the neighbourhood of C. , being the value of the modulus at C. In addition to the determinations of the periods of vibration , observations of the logarithmic decrement of the amplitudes of the oscillations were taken at each temperature , as already described .
These will be found among the results which follow .
At the end of the rigidity determinations for each wire , a series of observations was usually taken to ascertain the manner in which the logarithmic decrement and the torsional period varied with the amplitude of vibration , amplitudes up to about being used .
The main observations for the rigidity determinations were all taken at the constant average amplitude of 14 minutes .
Observations at 12 C. were only taken with a few of the wires , for great trouble was experienced through the heating jacket leaking when heated to this temperature .
In the account which follows , the complete set of curves and tables of results are given for a few only of the metals experimented on .
These serve as examples of the method of experimenting and as models of the results obtained .
It would be to remark here that only in a few instances was more than one specimen of the metal in question experimented on .
The values given for the rigidity modulus and its temperature coefficient would be only approximately true for any other specimen , for it is a well-established fact that the values of the rigidity modulus of any substance given by different specimens of the same material are not the same .
The cases in which more than one sample was experimented on were those in which the " " time effect\ldquo ; was least , as it seemed possible that in these cases the temperature ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
The values of the logarithmic decrements obtained at the various temperatures are given in Table IV .
Each is , in general , the mean of several observations .
TABLE IV.\mdash ; Soft Iron Wire .
From the numbers in the above table it will be seen that the logarithmic decrement decreases with increasing temperature .
This is just the opposite to what happened in the case of copper .
As a rule repeated heatings diminished the value of the logarithmic decrement .
This shows most clearly at the ordinary temperature , where the value of the logarithmic decrement is largest .
It is peculiar that after passing lo C. the value of the logarithmic decrement has attained a minimum and begins to increase with increasing temperature .
There is no doubt about this , for a large number of observations were taken at C. and at 12 C. , some of those at 10 C. being taken before and some after those at C. This result had also been obtained with another iron wire on which I had previously experimented , using the old form of heating jacket .
Another curious point is that the first heating increased the value of the logarithmic decrement at ordinary temperatures , while all subsequent heatings dimimshed it .
This result was also obtained with the first iron wire , and with the copper wire as can be seen in Table II .
In Diagram II .
a curve is drawn through the value of the logarithmic decrement from the first observation at each temperature .
When an the observations for finding the rigidity were finished , I investigated whether the time of vibration and the logarithmic decrement varied with the amplitude of swing so long as this was small .
The following two results were the means of several determinations : \mdash ; The increase in period at the larger amplitude corresponds to a decrease in the rigidity modulus of dynes per sq .
centim. The value of the logarithmic decrement is seen to increase a little with the amplitude .
The results of experiments on another iron wire , carried out in the old form of apparatus , the wire alone being surrounded by the heating jacket , gave as the modulus of rigidity at C. dynes per .
centim. , and for the " " temperature coefficient\ldquo ; at The wire used was exactly similar to the one the experiments on which have been described above , being a pure soft iron wire annealed in the same manner .
The large difference between the values of found for the two wires is most probably due to the imperfect method of experimenting used in this case .
Platinum .
A pure platinum wire was used .
The periods given in the following table are in all cases the means of several observations .
TABLE VI.\mdash ; Platinum Wire .
From the above table it will be noticed that , with the exception of the first heating , increase of temperature increases the logarithmic decrement .
This result is the opposite to that obtained for soft iron .
The logarithmic decrement was continually decreasing with the time , the rate of decrease being more rapid the higher the temperature .
This time effect masks the effect of temperature to such an extent that no useful knowledge is obtained by plotting the results given in Table VI .
When the determinations of the rigidity had been completed , a series of observations was taken to see in what manner the period and logarithmic decrement varied with the amplitude of vibration of the vibrator .
A set of observations was taken at temperatures very near to C. , and a second set at temperatures very near to C. The observed periods were corrected to these temperatures .
The results of these experiments are recorded on Diagram VI .
, in which the continuous lines show the alteration of the period of vibration , and the dotted lines the alteration in the logarithmic decrement of the oscillations .
It will be noticed that both the period and logarithmic decrement increase with the amplitude of vibration , and that the rate of increase of each is less at larger amplitudes than at smalIer ones .
The curves corresponding to the same temperature are of similar form .
A pure gold wire was used .
As with platinum and iron , the period of vibration diminished with time .
The amount of this diminution got less and less on repeated ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
logarithmic decrement increases as the temperature is raised , slowly at first , but more rapidly at higher temperatures .
This was especially noticeable after passing C. Experiments were made to ascertain the manner in which the period and logarithmic decrement varied with the amplitude of vibration of the vibrator .
Observations were taken at two temperatures , viz. , C. and C. It was found that the logarithmic decrement curves for the two temperatures were very much alike , and showed that the logarithmic decrement increases with the amplitude of vibration , the rate of increase being slightly greater at the higher temperature .
This is practically the same as the result obtained for platinum .
From the curves for the variation of period with the amplitude of vibration it was evident that the period increases with the amplitude , the rate of increase at the higher temperature being greater than at the lower .
Steel .
Two wires of pianoforte steel were experimented on , one in the first form apparatus in which the vibrator was not enclosed in the heating jacket , and the other with the apparatus in its final form .
Both were carefully annealed in the manner already described before the determinations of the rigidity were begun .
With the first wire the method of experimenting was slightly different from that used with the other wires .
The temperature in the heating jacket was regulated by means 01 a stream of water heated by passing through a gas furnace fitted with a quickly acting gas regulator .
By this means the temperature could be kept constant for any length of time , and the effect of very small alterations of temperature observed .
With this arrangement the temperature could not be raised above C. , but observations at C. and 10 C. were also taken by employing the method of heating described earlier in this paper .
I do not attach much importance to the numbers obtained , on account of the defective form of heating jacket employed , but the results can be relied on in so far as they show that the rigidity modulus of steel is a perfectly definite quantity , and that its rate of decrease with increasing temperature is constant also .
Table VII .
contains the values of the rigidity modulus at the various temperatures .
Each was calculated from the mean of a large number of observations ( in most cases more than 20 ) at the temperature given .
It must be stated that the results have been tabulated in descending order of temperature\mdash ; not in the order in which they were obtained .
This accounts for the fact that in two or three cases results are given at very nearly the same temperatures .
These were taken at different times , weeks apart , in order to see if the rigidity came back to its original value after the wire had been heated to a higher temperature .
TABLE VII.\mdash ; Steel Wire .
On plotting the values of the modulus of rigidity against the corresponding temperatures a straight line was obtained .
From this line the value of the modulus at C. was found by exterpolation and then the true temperature coefficient was determined .
The value found was , where .
The value of at C. was found to be dynes per sq .
centim. , from which , for the sake of comparison with the values for the other wires , the " " coefficient\ldquo ; was obtained , In determining the modulus of rigidity of the second steel wire with the later form of apparatus , use was made of the fact that it had already been found to be a linear function of the temperature .
Observations were therefore taken at two temperatures only , and from these the temperature coefficient was determined .
The following are the means of several determinations at the temperatures given .
From these the value of the modulus at C. was found to be dynes per .
centim. , and the " " coefficient\ldquo ; to be , The results of the experiments by which the moment of inertia of the vibrator was ON THE MODmUS OF TORSIONAL RIGIDITY OF METAL WIRES .
found with this wire will be found in Part I. of this paper .
Some experiments were made to determine the manner in which the period and logarithmic decrement varied with the amplitude of vibration of the vibrator .
The results are given in Table IX .
From these it is seen that the logarithmic decrement of the vibrations is less at the higher temperature than at the lower one .
This was also found to be the case with soft iron .
It is also seen that the period increases only very slightly with the amplitude so long as this is small .
The logarithmic decrements are slightly greater at the larger amplitudes , but the increase is not uniform .
Silver .
A pure silver wire was used , and it was found that the rigidity modulus for silver is much more nearly constant at a constant temperature than for most of the other metals examined .
This was particularly evident in the case of the observations taken at the ordinary temperature , which were not very seriously altered by the successive , especially in the case of those taken near the end of the There was , however , on the whole a slight increase of rigidity with time .
The rigidity-temperature curve for silver is given in Diagram VII .
The points marked represent the first and second observations at the ordinary temperature of the room and the means of the observations at the higher temperatures .
The line is slightly curved in the same direction as that for platinum ( see Diagram IV The value of the modulus of rigidity for silver at C. was found to be dynes per sq .
centim. , and of the " " temperature coefficient\ldquo ; The logarithmic decrements of the amplitudes of vibration when plotted against the corresponding temperatures gave a curve similar to that obtained for copper ( Diagram II Observations on the effect of the amplitude of vibration on the period and logarithmic decrement were made , as usual , at two temperatures , C. and C. It was found that\mdash ; as with most of the other wires examined\mdash ; both the period and logarithmic decrement increased with the amplitude of vibration , and that the rate of increase of each was greater at the higher temperature than at the lower one .
Oscillating the vibrator through a large amplitude was found to increase both the period and the logarithmic decrement at the smaler amplitude , indicating a diminution in the rigidity and an increase in the internal viscosity of the wire .
Aluminium .
The wire used was of pure aluminium , and the rigidity-temperature curve obtained was similar to that given by silver .
Several peculiarities were , however , noticed with this wire .
In the first place , heating to C. diminished the value of the rigidity modulus at C. , and this diminution was much larger after the wire had been heated to C. , but after heating to lo C. the rigidity at the laboratory temperature was increased , and this increase was continued after each heating to C. Successive heatings to C. sometimes increased and sometimes diminished the value of the rigidity modulus at that temperature .
It is also curious that , when kept at a constant temperature of C. or C. , the rigidity diminished slightly as time went on , but that at 10 C. it gradually increased with time .
It thus appears that the effect of heat on the rigidity of an aluminium wire is very irregular .
From the curve obtained , the modulus of rigidity for aluminium at C. was found to be ] dynes per sq .
centim. , and the value of the ' ' temperature coefficient\ldquo ; The logarithmic decrements of the amplitudes of the torsional vibrations were found to increase with the temperature , but the value at any one temperature decreased as time went on .
On plotting the logarithmic decrements against the corresponding temperatures , the curve drawn through the mean of the observations at each temperature was similar to that obtained for copper .
Observations made at larger amplitudes of vibration showed that both the period and logarithmic decrement increased with the amplitude , the increase being more rapid at C. than at C. The effect of swinging through a large amplitude was to lessen the value of the logarithmic decrement , and to increase the period of vibration at smaller amplitudes .
Tin .
A pure tin wire was used .
Owing to the great internal viscosity of this metal , the torsional periods could not be timed with the accuracy obtained in the case of The value of the rigidity modulus at C. was found to be dynes per sq .
centim. and the " " temperature coefficient\ldquo ; Some experiments were also made with another pure tin wire which had not been previously annealed .
The range of temperature employed in this case was only C. The value found for the rigidity modulus at C. was dynes per sq .
centim. and for the " " temperature coefficient\ldquo ; .
The rigidity of tin is thus lessened by annealing .
The " " temperature coefficient\ldquo ; is much the smaller in the case of the unannealed wire , but the alteration in the rigidity modulus per 1o rise of temperature is more nearly in the two cases .
The values of the logarithmic decrements of the amplitudes of oscillation obtained during the experiments with the annealed wire showed an increase in the logarithmic decrement with the temperature , and also a continual diminution with time .
Observations of the logarithmic decrement of the oscillations of the unannealed tin wire were made , and the results showed that the rate of increase of the logarithmic decrement with temperature was much greater than in the case of the annealed wire .
The heating of the unannealed wire greatly reduced the value of the logarithmic decrement at the ordinary laboratory temperature .
Some observations on the eflect of increased amplitude of vibration on the logarithmic decrement gay very irregular results , complicated by the fact that the logarithmic decrement of a constant amplitude diminished as time went on .
The observations showed that the logarithmic decrement increased with the amplitude of oscillation , but in a very irlegular manner .
Lead .
The wires used were drawn down from a strip cut from a sheet of lead .
In order to try to avoid the complications caused by the wire elongating under the slight tension to which it has to be subjected , series of observations at different temperatures were taken all on the same day .
Observations were first taken with an unannealed lead wire , and it was found that the rigidity increased very considerably as time went on , especially at the higher temperatures .
The rigidity-temperature curves obtain on different days were very similar , indicating that the alteration of rigidity with temperature was constant , although the absolute value increased as time went on .
A specimen of the curves obtained is given in Diagram VIII .
The value of the rigidity modulus at C. is dynes per sq .
centim. The mean value of the decrease per 1o rise of temperature at C. was dynes per sq .
centim. , from which the " " temperature coefficient\ldquo ; The logarithmic decrement of the amplitude of the torsional oscillations increased with the temperature , but this effect was complicated by a very considerable decrease in the value with time .
The results , however , showed that the rate of increase as ON THE MODULUS OF TORSIONAL RIGIDITY OF WIRES .
4 ] much greater at higher temperatures than at the ordinary temperature of the laboratory .
Some observations were made at the telnperatul'e of the ) oratory and at lo C. , on a lead wire which had been annealed .
It was found that the rigidity increased only very slowly with the time , and that the increase per 1o C. over this range of temperature was nearly the same as with the first wire .
This perhaps have been expected , for in the final of the modulus of the first wire a certain amount of annealing must have taken place owing to the wire having been kept for some time at 10 C. The increase of the modulus per 1o C. was for the annealed wire dynes per sq .
centim. , and for the unannealed wire dynes per sq .
centim. The value of the modulus at C. for the annealed wire was dynes per sq .
centim. Observations of the logarithmic decrement of the amplitudes of the torsional oscillations of this wire gave results similar to those obtained with the unannealed wire .
The wire was of pure cadmium , and was first annealed in the manner already described .
The small rectangular brass vibrator was used , but its weight caused the wire to elongate very considerably at the higher temperatures , and the results obtained at those temperatures were therefore unreliable .
A new wire was then used , and the temperature coefficient determined from observations at the temperature of the laboratory and at C. , the experiments being done as quickly as possible in order that the radius of the wire might be determined before it had drawn out to any serious extent .
The total elongation was 052 centim. , and a correction on this account was applied in determining the rigidity .
The value of the rigidity modulus at C. was found to be dynes per sq .
centim. , and the " " temperature coefficient\ldquo ; 005855 .
The logarithmic decrement of the amplitude of the torsional oscillations increased with the temperature , the rate of increase being much greater at 10 C. than at the ordinary laboratory temperature .
Copper .
An ordinary copper wire of the same size as the other wires used was carefully annealed in an atmosphere of carbonic acid gas , and a short series of observations taken with it , the object being to ascertain if there was any striking difference between the behaviour of this wire and that of the pure copper wire already described .
It was found that the rigidity at a constant temperature was far from being constant , in marked contrast to what was obtained with the pure copper wire .
On plotting a series of values of the modulus of rigidity against the corresponding VOL. CCIV.\mdash ; A. temperatures a curve , and not a straight line , was obtained .
The form of the curve was similar to that given by gold , iron , \amp ; c. , and suggests that the departure from the linear law is due to the gradual increase of rigidity with time .
The value of the modulus at C. for commercial copper is dynes per sq .
centim. , and the " " temperature coefficient\ldquo ; The modulus of rigidity is thus considerably less , and the temperature coefficient rather reater , than in the case of pure copper .
The logarithmic decrements of the amplitudes of oscillation were about four times as large as those obtained with the pure copper wire , indicating a much greater internal viscosity in the case of this commercial copper .
PART III .
SUMMARY AND COMPARISON 0F RESULTS .
The results of the experiments may be summarised as follows:\mdash ; 1 .
In all the materials examined , with the exception of pure copper and of steel , the modulus of rigidity at one temperature is not constant , but increases as time goes on .
The rate of increase of rigidity with time is greater the higher the temperature , and repeated heatings to the same temperature gradually lessen the rate of alteration with time at that temperature , but even in the course of months of experimenting the increase of rigidity with time cannot be entirely eliminated .
2 .
The diminution of the modulus of rigidity per degree rise of temperature between C. and 10 C. is constant for pure copper and for steel , but not for any of the other materials examined .
3 .
In the case of the metals iron , gold , tin , lead , and commercial copper , the rigidity-temperature curve is of such shape as to suggest that if it were possible to obtain the values of the rigidity modulus at the different temperatures , all within a very short space of ti-me ( so as to avoid the " " time effect the resulting curve would be a straight line .
4 .
In the case of the metals platinum , silver , and aluminium , the shape of the rigidity-temperature curve shows that the effect of the gradual increase of rigidity with time is such as to make the alteration with temperature approximate more closely to a linear law than would be the case if the observations were all taken within a very small interval of time .
For these metals , therefore , the decrease of torsional elasticity per 1o C. rise of mperature increases with the temperature .
5 .
In general , the effect of heating to a high temperature is to increase the value of the rigidity modulus at temperatures .
[ This applies even to pure copper , of which the modulus of rigidity at the ordinary laboratory temperature is slightly greater after the wire has been heated to higher temperatures .
The rigidity of steel ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
was quite constant , and with silver the value of the modulus at the temperature of the laboratory , in the last few experiments , was unaltered by a temporary increase of temperature .
The tin wire was the only case in which the rigidity modulus at the ordinary laboratory temperature was lessened by heatin ] 6 .
The internal viscosity of all the metals examined , with the exceptions of soft iron and steel , increases with the temperature .
This increase varies very much with different metals , being greatest with aluminium and least with platinum .
The internal viscosity of soft iron decreases rapidly with rise of temperature , and reaches a minimum value at about C. There is a slight decrease also in the case of steel .
7 .
Repeated heating and continual oscillation through small amplitudes decrease the internal friction .
8 .
Both the internal friction and the period of torsional vibration increase with the amplitude of oscillation .
The increase is generally greater the higher the temperature of the wire .
It is least in the case ) steel and is small in the case of soft iron .
9 .
Vibration through a large amplitude considerably alters both the logarithmic decrement and period of oscillation at smaller amplitudes .
The nature of the alteration varies with different metals , being in some cases an increase and in some a decrease .
10 .
The internal viscosity of a well annealed wire suspended and left to itself gradually decreases .
11 .
The internal viscosity of an unannealed wire is enormously reduced by annealing .
The values of the modulus of torsional rigidity obtained in the present research are collected gether in Table X. , which also contains the values given by the observers mentioned in Part I. of this paper , who have inyestigated the variations of the modulus with changes of temperature .
ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
The numbers in the above table show considerable discrepancies in the value found for the rigidity modulus of any one substance by different observers .
In many cases this is no doubt due to want of purity and to imperfect annealing of the specimens used , also perhaps in some cases to the fact , that the wire was subjected to excessive tension during the rigidity determinations .
Owing to the fact that different observers have expressed the variation of rigidity with temperature by different formulae ( already quoted in Part I. of this paper ) , it is not easy to give a comparison of their results .
It appeared to be best to calculate in each case the average decrease in rigidity per C. rise of temperature per unit value of the modulus over the temperature C. to lo C. , this being the most generally employed range of temperatures .
In order to obtain numbers from the present research to compare with those of other observers , a temperature coefficient was calculated for each wire from the last set of observations at 10 C. , and at the ordinary laboratory temperature , assuming the rate of alteration of rigidity with temperature to be constant in the interval .
These numbers are thus the values of in the equation The coefficient \mdash ; the alteration per ' 1o C. rise of temperature per unit value of the rigidity modulus , at C.\mdash ; are also included in Table XI .
ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
The values given in the above table were not in all cases found from the range of temperatures C. to 10 C. The experiments of NAPIERSKY and those of KUPFFER on iron , platinum , and silver , were only performed over a small range of temperature , from about C. to C. , while the numbers of SCHRFER and BENTON apply to temperatures between C. and C. It must also be pointed out that KOHLRAUSCH .
and LOOMIS , PISATI and TOMLINSON , give formulae showing that the alteration of rigidity with temperature does not follow a linear law ; KOHLRAUSCH and LOOMIS , and TOMLINSON , finding that the rate of decrease per degree rise of temperature increases with the temperature , a view with which PISATI'S results , with a few exceptions , agree .
The values given by KOHLRAUSCH and LOOMIS are very different from those obtained by other observers , a result no doubt due in part to the fact that the wires they used had not been subjected to any annealing process .
These authors state that observations taken on different days could not be compared , and their tables it is seen that the period of torsional oscillation decreased from day to day .
This means an increase of rigidity with time , effect to which I have drawn special attention in this paper , and which , in unannealed wires , is quite large .
A fault common to all the experiments which have hitherto been made by the dynamical method , is that sufficient precautions were not taken to insure that during the observations at high temperatures the wire was everywhere throughout its length at the same temperature .
The average temperature of the wire must generally have been considerably below the temperature indicated by the thermometers .
This was found to be the case when experimenting with the first form of apparatus used in the present research .
Under these circumstances , the observed rate of alteration of rigidity with temperature would be less than the actual alteration , and the temperature coefficients obtained would be too small .
In the present research the values of the temperature coefficients for iron and steel , as given by experiments with the old form of apparatus , are less than the values obtained with the apparatus in its final form .
The results of KUPFFER and of NAPIERSKY are only of interest from an historical point of view , they differ very considerably from those obtained by more recent observers .
One great fault in their experiments is that the amplitudes of oscillation were extremely large .
In experiments , too , the amplitude of vibration was at the beginning of each observation , and the rigidity of the wire ( the length of which was 65 centims. ) must have been considerably altered by being subjected to this treatment .
Of the more recent experiments , those of SCHfflFER stand out as comprising the largest number of materials .
As already stated , he employed the statical method of experimenting , observing the torsional deflections of a wire under a constant couple .
The temperatures worked at were , in general , only two , the temperature of the laboratory and that of boiling liquid air , and the rate of alteration of rigidity with temperature was assumed to be constant between them .
The main object of this work was , however , to compare the temperature coefficient of the torsion modulus with the coefficient of expansion , melting-point , and atomic weight of the metal in question , rathel than to obtain great accuracy in the investig tion of the effects of .alteration of temperature on the rigidity modulus .
vations of the logarithmic decl.ement of the amplitude of the torsional oscillations were not the main object of this research , the results obtained throw considerable light on the internal viscosity of the wires used , and it seems desirable , therefore , to compare them with similar observations by other observers .
The diminution of amplitude of the torsional vibrations of a wire suspended as in the present experiments is due to two causes:\mdash ; ( 1 ) loss of energy by the vibrating system to the surrounding air ; ( 2 ) the internal molecular friction of the suspended wire .
In order , therefore , to obtain accurate information as to the internal friction of the wire , it is necessary to deduct from the observed logarithmic decrement that portion which is due to the viscosity of the air .
An exact calculation of this is impossible in the present case on account of the various objects fixed inside the box enclosing the vibrator , but an approximate value was found by using MAXWELL'S formula for the case of a disc vibrating between two fixed plates .
The value thus found for the logarithmic decrement due to air-damping was of the order .
The increase in this value due to an increase of the temperature of the air of C. ( about 25 per cent. ) would not be within the limits of accuracy of the observations , and hence the variation in the observed logarithmic decrement on increasing the temperature of the jacket enclosing the vibrating system may be taken as being due entirely to the alteration of the internal viscosity of the suspendin wire .
Observations of the effects of changes of temperature on the internal viscosity of metal wires have been made by most of the observers who have investigated the modulus of rigidity by the dynamical method , notably by KUPFFER , PISATI , and TOMLINSON .
KUPFFER was probably the first to observe that the rate of subsidence of torsional oscillations increased with the temperature of the wire .
He did not investigate this effect fully , merely stating that at each temperature the amplitudes of successive oscillations decrease in geometrical proportion , a conclusion which is not borne out by the numbers given .
PISATI , who experimented over a range of temperatures from C. to 30 C. , investigated this effect more fully , and gives a series of formulae for calculating the amplitude of the oscillation after the amplitude was , ovel different ranges of temperatures .
He found a marked increase in the of subsidence at higher temperatures except in the case of iron , in which the number of oscillations made by the wire in damping down from an amplitude of to an amplitude of at first increased with rise of temperature and reached a maximum at C. , afterwards decreasing again and becoming constant between 20 C. and This minimum internal viscosity for iron at ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
C. was also observe in the present research , but it is not mentioned by many experimenters .
STREINTZ* was the first to lay stress on the fact that there is a tendency in metallic wires to accommodate themselves to a definite elastic state .
In wires of brass , silver , platinum , copper , iron , and steel examined by him there was a marked diminution in the rate of subsidence produced by keeping the wire for long periods in continual torsional oscillation .
Lord KELVIN had previously stated that the rates of subsidence of the vibrations of several wires tested by his students were much less rapid after they had been allowed to rest for a week-end than when kept in continual vibration .
The present research shows that the whole question depends on the amplitude of the vibration .
It is possible that the wires mentioned by Lord KELVIN ( which had been kept " " as continually as possible in a state of vibration presumably by impulses administered by students ' hands ) had been subjected to larger amplitudes of oscillation than were used in the case of the wires which were allowed to rest between the successive observations .
In my own experiments on the effect of the amplitude of vibration , in which the greatest amplitude used was about or , it is shown that , in general , vibration through a large amplitude very considerably alters both the period and logarithmic decrement of the oscillations at smaller amplitudes .
Recently investigated this question and arrived the conclusion that , when the temperature and amplitude of vibration are constant , both the period and logarithmic decrement are constant quantities , but his experiments do not seem to be capable of sufficient accuracy to establish the point .
STREINTZ expressed the results of his experiments in the following three laws , the first of which is not strictly borne ont by his own results:\mdash ; .
The logarit , hmic decrement is independent of the amplitude of vibration , of the tension in the wire , of the period of oscillation , of the perhaps also of the diameter of the wire . .
The logarithmic decrement varies with the chemical and physical nature of the wire .
It increases with rise of temperature and is less in annealed than in unannealed wires . .
The logarithmic decrement tends\mdash ; up to a certain limit\mdash ; to diminish with the number of oscillations performed by the wire .
Perhaps the most exhaustive experiments on the internal viscosity of wires have been made by TOMLINSON .
S He obtained a decrease of the logarithmic decrement in the case of iron , nickel , and platinum , on raising the temperature from C. to the percentage decrease being much the greatest in the case of iron .
He states that * STREINTZ , ' Sitzb . .
Wien .
' LXIX .
, Abtb .
2 , p. 337 , 1874 ; also LXXX .
, ) , p. 397 , 1880 .
THOIISON , 'Proc .
Roy .
Soc vol. 14 , , 1865 .
THOMPSON , 'Phys .
Rev VIII .
, p. 141 , 1899 .
S TOULINSON , ' Phil. Trans vol. 177 , p. 801 , 1886 ; also ' Proc. Roy .
Soc vol. 40 , 1 ) .
343 , 1886 .
VOL. CCIV.\mdash ; A. if we start with a sufficiently low temperature the internal ffiction of all annealed metals is first temporarily decreased by rise of temperature and afterwards increased .
The temperature of minimum internal friction is for most annealed metals between C. and In the present research this effect was obtained in two cases only , , for soft iron and for steel .
There was , however , a slight decrease in the value of the logarithmic decrement at the first rise of temperature in the case of platinum and of commercial copper , but this was , I think , due to the " " time effect\ldquo ; ( a decrease of internal viscosity with time ) being greater than the increase due to the rise of temperature .
This time effect diminishes as time goes on and the decrease of logarithmic decrement with increasing temperature did not occur again .
TOMLINSON'S other results are practically the same as those of STREINTZ , mentioned above .
He , too , states that the logarithmic decrement of the torsional oscillations is independent of the amplitude of vibration , provided the deformations produced do not exceed a certain limit , varying with the nature of the metal .
This is contrary to the results obtained in the experiments now described , in which it was found that the logarithmic decrement increased as the amplitude increased from 14 ' to In some cases the rate of increase was greatest at the smallest amplitudes , and it may be that in these cases , over larger amplitudes , the logarithmic decrement remains constant , but this is not a general law .
TOMLINSON also states that arithmic decrement of the amplitudes increases with the period of vibration , on this point differing from .
In the pIesent research it was found that the period increases with the amplitude of vibration .
If , then , TOMLINSON is right , it would be expected that the logarithmic decrement would also increase with the amplitude of vibration , and not , as he states , be independent of it .
PART IV .
THE DETERMINATION 0F COEFFICIENTS 0F LINEAR EXPANSION .
The coefficients of expansion of the metal wires , and also of the gun-metal bar cast at the same time as the ring used in determining the moment of inertia of the vibrator , were rmined by means of the measuring bench in the Physical Laboratory of the University of Birmingham .
This ] lent instrument was designed by Professor pointing , and he very kindly allowed me to work with it during the Cambridge vacations .
As no description of this measuring bench has yet been published , I will describe it in detail , as well as giving the special arrangements adopted for measuring the expansion of wires .
A diagram of the apparatus is given in fig. 4 .
It consists of two microscopes held rigidly in vertical positions by massive iron stands , which rest on a slate window-slab and can be placed at any distance apart .
The microscopes project over a ON THE MODULUS OF TORSIONAL RIGIDITY OF METAL WIRES .
railway , along which two troughs , one containing the article under test and the other a standard metre , can be moved .
The microscopes ( iig .
4 ) are each fitted with a rocking-plate micromete ' by means of which measurements to .
can be made .
The trough in which the article to be measured is placed is a copper-lined box about 120 centims. long by 18 centims. high and 10 centims. deep .
It is divided into two halves by a partition down the middle , stopping short at about 5 centims. from each end : This serves to direct the circulation of the water with which the trough is filled .
The article under test lies on the side of this partition nearer to the slab on which the microscopes are supported .
The arrangement may be seen in plan in fig. 5 .
The water is by means of a fan ( fig. 5 ) worked by a motor ( fig. 4 ) .
The fan revolves in a box formin a depression in the floor of the trough at its middle .
It is situated on the opposite side of the central partition to that occupied by the article whose expansion is to be determined , and drives the water up on that side , thus causing a circulation in the direction shown by the arrows in fig. 5 .
The trough containing the standard metre is exactly similar to the one just described .
The metre is supported in a horizontal position on four centims. apart .
The first and second , and the third and fourth of these connected together by light levers having fulcrums at their centres , which are supported from the bottom of the trough .
This method allows of the metre being supported without bending and also leaves it quite free to expand .
* For description , see POINTING , 'Phil .
Trans , vol. 182 , p. 589 , 1891 , or " " The Density of the Earth p. 95 .
The arrangements for holding the wire are best seen in fig. 6 .
One end , , is firmly clamped and the other is attached at to a vertical lever CBD , of which the fulcrum is at C. A string is ttached to the lever at , and this sses down the grooved side of the metal piece EFG ( which serves the same purpose as a frictionless pulleywheel ) and carries at its lower end the mass H. mass is so chosen that the tension produced in the wire is the same as that due to the weight of the vibrator in the rigidity experiments .
The wire is surrounded by water except at the points the distance apart of which is to be measured .
At these points it passes transversely through small brass tubes closed at the lower ends .
These tubes are supported in a vertical position by wires from the base bar L. The holes in the tubes were made the same size as the cross-section of the wire and just outside , both wire and tube were covered with melted shellac , which effectually prevented any water leaking in .
The standald thermometer was supported beside the wire by means of supports from the base plate L. The water in the trough can be heated by means of the row of burners , shown in fig. 4 .
The gas feeding these passes through a gas regulator , the thermal part of which is a mercury bulb , centims. in diameter , running the whole length of the trough on the opposite side of the central partition to the wire .
With this heating and stirring arrangement the water in the trough can be kept at a temperature not varying by more than C. for hours .
The top of the trough was covered over with pieces of bright tin plate , a small circular hole being left under each of the microscopes .
These holes were also covered over except when readings were being taken .
The thermometer was read to C. by means of a microscope , the covering of the trough above the part of the stem which was required to be seen being removed for this purpose .
In order that the rocking plates and objectives of the microscopes might not be clouded with steam at the higher temperatures , a layer of ordinary lubricating oil was poured over the surface of the water .
This effectually prevented any steam escaping , and also helped in maintaining the temperature constant .
The microscopes were arranged on the slab so that each was over the corresponding tube on the wire , and particular care was taken to see that the wire was horizontal and parallel to the direction of motion of the trough .
The height of the second trough containing the standard metre ivas adjusted so that the scale could be seen clearly without re-focusing the microscopes .
The difficulty in measuring small alterations in length directly by means of a microscope is to obtain a sufficiently fine fiducial mark .
This difficulty is especially great in the case of a wire , on which it is veIy difficult to make a fine mark .
After many failures it was found that the best plan was to map out the appearance of the surface of the wire and fix on some easily distinguishable point .
The illumination was secured by an excellent arrangement designed by Professor POINTING .
By means of this the light from a small 10-volt electric lamp , enclosed at the end of ON THE MODULUS OF TORSIONAL IDlTY OF METAL WIRES .
a horizontal metal tube some 20 centims. long , can be focussed on to the wire , being reflected downwards by a ngled prism .
The path of reflected beam is almost vertical and is in a perpendicular to the wire .
In this manner perfect illumination is secured , and when looking at the standard metre no shadows are seen along the graduation lines .
In determining the coefficient of expansion of the gunmetal bar it was supported in a slnlilar manner to the standard metre .
The fiducial marks were near the ends of the bar and were made by brightly polishing the surface and then drawing a fine needle across .
The marks protected from the water by two glass ) which had been ground level at the bottom and were held firmly against the bar by stout elastic bands , a ring of thin sheet indiarubber being placed between .
In experimenting , the trough containing the wire under test was left under the microscopes only while readings were being taken .
It was moved off along the line to be heated up .
At each temperature about half-an-hour was allowed after the required temperature had been obtained before the readings were taken .
Durin this period the heating and stirrin were continual .
The trough was then wheeled under the microscopes and adjusted in position , so that the marks were approximately under the cross wires .
The heating and stirring were then stopped\mdash ; for the shaking rendered the image indistinct\mdash ; and by means of the rockin plates the marks were adjusted on to the cross wires .
The temperature was then noticed and the trough wheeled away .
The readings of the micrometer scales were then taken .
This was repeated three or four times at each temperature , and in general there was no difficulty in obtaining a very close agreement .
The expansions were taken at various temperatures up to C. , and after each observation the distance between the microscopes was checked by wheeling the standard metre them .
This precaution was necessary in ordel to see that the heat did not warp the stands and thus alter the positions of the microscopes , but only in a few cases was it necessary to apply a correction on this account .
The coefficients of expansion of all the wires used in the rig.idity were determined , with the exception of the very soft metals lead , cadmium , and tin , which were found to draw out slightly under the necessal.y tension , especially at the higher temperatures .
The coefficients of expansion of these metals would have to be from bars of the material , but in the present research it was not necessary to know them with great accuracy on account of the fact that the rigidity determinations in these cases were not so accurate as with the harder wires .
The values in LUPTON'S tables .
( 1901 ) were therefore used .
Before determining the coefficients of expansion , the wires\mdash ; which had been bent into spirals for weighing in air and water in the determination of their radii\mdash ; were straightened and annealed in the same manner as before the determination of their rigidities .
The lengths of the wires.at the different temperatures were plotted against the corresponding temperatures , and in every case the resulting curve was a straight line .
lengths of the wires at C. were found by exterpolation and the values of the coeflicients of expansion calculated .
In order to show the accuracy obtained , the lengths of the wires at the different temperatures were calculated from the formula , where and are the length of the wire at C. and C. respectively and is the coefficient of expansion .
It is peculiar that in several cases the difference between the observed and calculated values was far larger at the ordinary laboratory temperature than at any other .
Below are given the complete set of observations on the first wire tested , viz. , soft .
For the other metals I give merely the value found for the coefficient of expansion and the average difference between the observed and calculated lengths at the various temperatures .
For the observations recorded in Table XII .
the coefficient of expansion for soft iron is 00001118 .
Another determination gave .
The mean of these , , is the value used to correct the observed periods of oscillation in the determination of the temperature coefficient of the rigidity modulus .
The coefficients of expansion of the other wires experimented on and also of the gun-metal bar already referred to are contained in Table XIII .

