]\gt ; XIII .
] O. W. TCHARDSON , , B. Sc. , of Trinity College , .
icated by ofcssor J. J. THOMSON , F.R. eceived , \mdash ; Pead , 1903 .
THE experimental part of the present paper is an investigation of the electrical conductivity of the space surrounding hot surfaces of platinum , carbon , and sodium at low pressures .
A preliminary account of some of the experiments on platinum was read before the Calnbridge Philosophical Society on Noyember 25th , 1901 The conductivity produced by hot metals has been the subject of a great number of researches by diflerent authors .
The phenomena are , however , very complicated ; for the quantity and sign of the ionisatio1l is found to vary in the most remarkable manner with the nature , temperature , and previous history of the metal , with the nature and pressure of the surrounding gas , and with small changes in the state of the metal surface .
The present investigation was undertaken with the idea that the ionisation at high telnperatures the conductivity ) roduced by metals took its simplest form .
This idea ) is supported by the observation of Professor that the } tive current is to a extent independent of the nature of the gas , and is indel ) endent of its pressure ange from 04 to 004 millim. The chief problem which is attacked ) erimentally is the in the saturation current from the hot metal surface to a neighbouring electrode varies with the metal of the satnration current corresponds to the total number of ions which are produced by the surface per second .
it was found necessary to examine , in addition to the aboye , the relation between the current and the electromotive force for the conductivity produced by the three abovementioned conductors at various pressures .
The theory , by which it is proposed to explain the phenomena , is based on the ' Proc. ) .
Phil. vol. 11 , p. 286 .
'Proc . .
Phil. ' vol. 10 , p. 241 , , p. 296 .
VOL. 20.7.03 conduction metals by corpuscles which has been elaborated by Professors and J. J. THOMSON .
T According to that theory a metal is to be sponge-like of molecules and comparatively large fixed positive ions , ith small negative ions or COl.puscles l1loving freely with great velocity throughout the .
Since the COl.puscles do not all leave the metal when they strike the , it is evident that there must be a Stlt.face discontinuity of potential escal ) we xise the temperature of the metal we the velocity the , provided the energy equired to surface ( ] with the ture , 1Y1 ny more the , which the will through than before .
In this way we calt in the number corpuscles shot off from unit area of a metal .
The folmulas ) ined involves two new constants , , the numbel of ions in unit volume of metal and the work done by an ion in through the surface .
It to state in that ( the of the results in iking agreement the theory .
In particular the etical f makes turatiol 1 current mously rapidly with the of it exponential .
The exlJel.iments show that this actually tion current ovel the following Ilductol examined : to I per centinl .
2 .
, , , , , , total current .
The anges teml)erature for platinum and roughly from 1000o C. to 1600o C. , from 10 C. to 45 C. respectively .
haps t most surprising esult of the investigation is the relatively enormous curl.ents which have been obtained .
The biggest leak measured was ampere from a to an electrode placed neal it ; this responded to a current of 2 per sq .
centim. of the carbon surface , the potential on the being volts .
In this case the pressure was only millim. of mercury , so that the ionisation produced by collisions was negligible .
In cases care was taken that the field which was put on the filaments was insufficient either to start a or to lnaintain one when started .
The naller currents with sodium were measured by means of a quadrant electrometer : the largest ampe was registered on a Weston ammeter .
It is evident from these experiments that a metal if placed in a yacuum and heated to a sufficiently high temperature makes the space around it an extremely ' vol. 1 , p. 566 .
do Physique , ' , 1900 .
IMPARTED TO A VACUUM BY HOT CONDUCTORS .
good ) of electricity .
The results that in the descent , heated to the ] rhestture it ill stand , the specific conductivity of the .
space is with l. the filament .
collduclol the to the will , of coul.se , only go in one ection .
heJl t is charged positively , and the put to ] , is small in with the current when the wire is tively .
The inder of the present is divided A. I. Calculationl of the lion c II .
of corpuscles neal hot plane of finite a B. I. Experiments with platinum .
II .
, , , , carbon .
III .
, , , , sodium .
C. clusion . .
I. of the ration Cnrrent S 1 .
The application of the kinetic theory of gases to the equilibrium ) the free negative electrons or corl ) uscles inside a metal scarcely needs justification here , since it has already been made use of by Professol DRUDE .
, It may , however , be permissible to point out some results which show that the similarity } ) etween a corpuscle in a metal and a molecule in a gas under ordinary conditions is close indeed .
Professor THOMSON has shown , from the change of resistance of bismuth in a magnetic field , that the mean path of a corpuscle in that metal has the value 1 centim. ; while a serieH of experiments by Mr. PATTERSON indicate that platinum , gold , , silver , copper , zinc , cadminm , me ] , Rlld carboll the lleall has values lie between and centinn .
The mean free for a ] ogen molecule air under conditions is centim. ; so that the mean free path corpuscle in bismuth is ) as of a molecule in air at ) osphere pressure , whereas other nletals the meall free path is the saYlle as tha in air } about tmospll e pressure .
free is , ' vol. 1 , ) .
' Rapports presents an de 'Phil .
Mag. ' ( 6 ) , vo1 .
.3 , 1 ) .
655 .
:3 W. RTCIIARDSON ON , ELECTRTCAL CONDUCTIVITy of course , only one-hundredth that of air molecule possessing the same free path , owing to the great velocity of agitation of the corpuscles .
Neyertheless we are quite justified in assuming that the time during which the corpuscles are moving freely is great compared with that during which they are colliding .
In fact this assumption follows at once if we are to attach any definite meaning to the ideas of free ths and collisions .
If , in addition to neglecting the number of corpuscles which colliding at any moment in comparison with those which are not , assume that the atoms of the metal and the itive ions oscillate about fixed centres and are subject to of restitution which are functions of their displacemelts only , we obtain at once , by the application of the ordinary analysis of the kinetic theory , the distribution of velocity among the corpuscles .
This is found to be the same as that for an equal number of similarly constituted gas molecules .
Thus the number of corpuscles which have velocity components in three mutually perpendicular directions between ?
and and , and and respectively are given by and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and ( 1 ) , where is the total number of corpuscles considered , is the mass of a corpuscle , , and are the impressed velocity components of the corpuscles in the direction of , and respectively , whilst is the average energy of translation of a corpuscle , and is equal to that of a gas molecule at the same temperature as the metal considered .
The velocities are connected with the components of the current density according to the relation where is the charge on a corpuscle , is the number per cub. centim. S2 .
If we suppose the impressed velocities to be nil or to be negligible } ) ared with the velocities of agitation , the number of cules in unit volume having velocity components between and and ) and ?
) ' becomes . . . . . .
( 2 ) , whilst number with these velocity components which hit unit area ) endicular to per second is . . . . . .
( 3 ) .
TO A VACUUM BY HOT CONDUCTORS .
If we suppose the of the ] conductor to } ) perpendicular the axis of ?
then total lnmbe of corpuscles which hit unit of the surface ) ' second is We now suppose that there is a discontinuity in the ) at the .
ce of the lnetal which is enough to the of the COl.puscles at low .
If the work -one by ion in through the surface layer is , then the tinuity in the potential is , where is the charge on an ion .
We have ful .
there , luy symmetry , the surface being perpendicular to the axis of Moreover ? ?
, whence where is the normal component of the velocity of the corpuscle ftel .
it has escaped from the metal .
It is evident from this that not all the corpuscles which strike the surface of the metal escape from it , but only those which have a normal velocity component which is .
Hence , to get the total number which through the surface layer , have to integrate expression ( 3 ) with respect to not from to , but from to .
Thus the total number which escape per second from unit is iyen by ( 5 ) ( b ) , since is connected with , the absolute telnperature , by the 1 being the gas constant for a corpuscle , The saturation current being equal to the quantity of electricity carried by the ions which are shot off from the surface in one second , is given by ' . . . . .
( 7 ) , where is the ) the nletal surface and , the on iou .
3 .
When the ions are removed by ) electric field as quickly as they IIR .
O. W. RICHARI)SON ON THE ELECTRICAL CONDUCTIyITy set free the surface of the metal , as in the case of the to be described later , the ] must be continually losing energy owing to the emission of the corpuscles .
This energy is composed of two parts : the first , represented by the done by the corpuscles in passing through the surface layer , while the second is equal to the ellergy of translation which they possess when they have reached the outside of the metal .
The sum of the two is easily calculated , since it is to the enel.gy of translation which the corpuscles that have passed through the layer possessed while they were inside the metal .
We have therefore nlerely to multiply the number of corpuscles which strike the surface by the energy each ossesses and grate between limits which embrace all values that pass through the layer .
The total loss of energy per second is therefore ( S ) , ( 9 ) .
Now the work done in a second by the corpuscles passing through the surface layer is obviously , so that the part of the energy lost by the hot metal per second .
which appears in the form of the translational energy of the corpuscles , is given by . . . . .
( 10 ) .
This calculation of the rate of emission of energy only applies , of course , to the case where the ions are remoyed by an external field as fast as they are formed .
If there is no external field and the ions are allowed to remain , we soon arrive at a steady state when as many corpuscles possessing a given amount of energy enter the surface of the metal in a given time as leave it ; so that in this case there is no loss of energy due to this cause .
The following proof of formula ( 6 ) , which is due to Professor J. J. , is interesting , since it does not involve the methods of kinetic theory of gases .
Suppose we have a closed space which is bounded by a surface of hot metal , then the corpuscles will be given off from the metal until a steady state is reached .
In this steady state as many will pass through the bounding surface from the acuum to the metal as from the metal to the vacuum , but the pressure will not be the same on both sides of the surface owing to the forces which tend to retain the corpuscles in the metal .
There will thus be a discontinuity in the pressure at the surface of separation , and being the work done on all ion when taken through the surface , we have , TO A yACUUM BY HOT CONDUCTORS .
.503 where 1 refers to the metal and 2 to the neighbouring space , being the pressure the volume occupied a corpuscle at any poin .
Substituting its value from equation , we get ; whence , , be the ) .
of corpuscles per unit volume outside and the number per unit inside the metal , we Now the 1lulnber of } ) uscles shot off the surface per second is not equal to the number per unit volume of space , but is equal to this multiplied by the velocity perpendicular to the surface .
So that , in the steady state , , where ; whence hicb is As ) eeIl deduced without postulating the existence of a steady state .
By following up the analogy betweell the e1llissiolt of corlJuscles and evapol.ation , the preceding formulae , connecting the ) with the , can be hermodyIlalnically in a manner still fewer assumptions .
II .
The of of Hot Metal of Infinite Extent .
S4 .
Both problem and the corresponding problem in spheres are of considerable portance , not only in connection with experiments in vacuum tubes , but also with regard to the behaviour of hot celestial bodies in space .
For instance , the aurora borealis and allied phenomena indicate that large quantities of ions continually reach the earth from some extraneous source , while certain variations of the earth 's magnetic field and other meteorological phellomena seem to be intimately connected with events which take place at the surface of the sun .
The present paper does not attempt to solve these questions , but the above facts indicate that the subject of the ionisation produced by hot bodies is not without importance in regard to The problem under consideration may be specified in the following terms :Given an infinite quantity of hot metal bounded on one side by a plane surface of infinite extent which is maintained at a given potential , find the charge on unit area of the metal surface and the potential at any point in the space outside the metal when the steady state has been attained .
Let us take the surface of separation perpendicular to the axis of , and let the suffix 1 refer to points inside the metal , the suffix 2 referring points in the ,504 .
O. W. ON CONDUCTTVITY neighbouring empty space .
There is , as we have seen already , a discontinuity in the pressure the corpuscles the surface of the metal , and by the conservation of ( 11 ) , where is the work done in -mass of ) corpuscles through the sul.face layer , is the pressure , and the volume of unit mass of the corpuscles at any point .
Similarly , in to btain the equations satisfied by the corpuscles outside the metal when the equilibrium stage has been reached , we use the principle that the work along any path extending from a point to a point due to expansion is equal to the work done by the electric .
This gives / V being the electrostatic potential , , the charge on a corpuscle , and the number of corpuscles in unit volume , since everything is independent of and .
Now the number of corpuscles in unit mass , whence .
( 12 ) .
In addition to this the electrostatic potential has to satisfy POISSON'S equation , takes the , being the numerical value of the negative charge .
The equation to be satisfied is therefore , where ( 1.3 ) .
A first integral of this equation is an integration constant .
Now when is infinite , and therefore , so that IMPARTED A VACUU BY HoT CONDUCTORS .
whence A being a second integration constant .
We have the further conditions : whenwhen which are satisfied if Taking the positive root for ( since negative values are inadmissible ) and putting we obtain . . . . .
( 14 ) .
This equation gives the concentration of the corpuscles at any distance from the plane when the temperature is maintained at absolute .
Returning to equation ( 12 ) we see that integration and substitution for yield the electrostatic potential V in the form If for , the integration constant is determined as so that V is finally to be obtained from ( 15 ) .
The electric intensity at any point is given by ( 16 ) , and the charge on unit area of the radiating plane by ( 17 ) , VOL. CCI.\mdash ; A. the volume density at any point bein \mdash ; \mdash ; ( 18 ) .
It is evident that , since for .
Thus , as we should expect , the charge on the surface is equal and opposite to the total charge in the space outside the metal .
As a numerical illustration we may calculate the potential at a point distant 10 centims. from a plane surface of platinum which is put to earth and maintained at a temperature of , say , absolute .
Taking the number of molecules in a cubic centimetre at and 760 millims. as , the charge on an ion as and the value of , which has been determined experimentally , to be we find the potential at a point 10 centims. from the surface to be volts , while at a point 1 centiln .
distant it would be about volts .
The experiments in the were not intended to test this part of the theory , but they show , as we should expect , that practically the whole of the current is stopped by a fall of potential of the order of one volt when it tends to drive the COl.puscles back to the hot metal .
It will be seen by inspection of formula ( 15 ) that even at the highest temperatures we can attain the potential differences at small distances from the hot surface never become very great .
For instance , at the temperature of the sun ( 6000o C. ) the difference of potential between the surface and a point 1 centim. distant from it would be only about sixteen times its value at 1300o C. On the other hand , the surface density increases very quickly with the temperature , as will be seen from formula ( 17 ) .
In the case of carbon at 6000o C. , taking as a probable maximum value of and as the value of , we find that has the enormous value of 300 electrostatic units , whereas at 1300o C. would have been less than this in the ratio of 1 to These numbers are to be taken as purely illustrative .
It is not supposed that any conductor could possibly exist in a vacuum at 6000o C. It will be noticed that the preceding theory of the equilibrium of corpuscles near a surface where they are being emitted is quite independent of any hypothesis as to the nature of the mechanism by which they are set free .
The results are therefore of interest even if the hypothesis , that the negative ions from hot conductors are the same as those which carry the current inside the metal , is ultimately found to be untrue .
IMPARTLD TO A VACUUbI BY HOT CONDUCTORS .
xPERIMENTAL INVESTIGATION .
I. Expcr.iments with S 1 .
Description of the The ultimate object of the experiments was to determine the way in which the saturation current from a hot platinum wire to a surrounding electrode , both placed in a vacuum , varied with the temperature of the wire .
For this purpose the type of bulb shown in fig. 1 was found to be most convenient .
The wire to be heated was in the form of a spiral , with its axis passing centrally along the length of the tube ; the current through the wire was supplied by means of the two thick leads and .
The electrode to which the current was measured was an aluminium cylinder which surrounded the hot part of the wire .
The cylinder was supported by a stout aluminium wire , sealed through the side tube by means of platinum .
The end was connected to the electrometer or galvanometer which served to measure the current .
The side tube connected the bulb with the pump and McLeod gauge .
In the earlier experiments , trouble was experienced owing to loose contacts appearing at and when the platinum wire had been heated .
In the { inal form of the tube this was obviated by making the leads and of platinum wire 1 millim. thick , to which the ends of the platinum spiral were welded electrically .
This made the platinum quite continuous through the tube .
The support of the electrode was insulated outside the tube at by means of sealing wax .
Inside the tube there was only glass insulation , which , however , is very good at low pressures .
The temperature of the platinum wire was obtained by measuring its resistance .
The arrangement of apparatus which was used to do this and to measure the current from the surface of the wire is indicated in fig. 2 .
The whole of the apparatus below is the part which was used to determine the resistance and was insulated on paraffin blocks .
It could be charged to any desired potential up to 400 volts by means of the battery through the key and water resistance A. The potential was measured by the volt-meter W. In this way any desired potential could be maintained on the hot wire F. The cylindrical was put to earth through the galyanometer , which thus served to measure the current .
In some of the experiments a Thomson galvanometer , of 4058 ohms resistanoe , which gave a deflection of 1 scale division for ampere , was used , but this was replaced later by a less sensitive D'Arsonval galvanometer .
The sensitiveness of the galvanometer deflections could be lowered at will by means of the shunt resistances , and The resistance of the hot wire was determined by placing it in one arm of a Wheatstone 's bridge , the other arms being the resistances , and The battery , which worked the bridge , also supplied the current necessary to heat F. As there was a current of up to amperes continually running through the arm , the corresponding arm had to be constructed so as to carry this current without heating .
A erman silver heostat of very thick wire mounted on an iron frame and having a resistance of ohms was found to fulfil the required conditions .
The arm had a resistance of 1000 ohms , and plugs were taken out of the box till a balance was obtained in the galvanometer .
The maximum current which flowed through the arms and was thus about of that flowing through and F. There was therefore no danger of and being heated by the currents used .
The adjustable resistance served to regulate the heating current .
By means of the keys , , the battery could be put on between the two resistances and at the one end , and and at the other .
In this way the currents flowing through the various arms could be made small , would not become heated , and , by making the proper connections for the galvanometer , the resistance of the arm at the ordinary temperature could be determined .
In the earlier expel.iments an arrangement was used which enabled the same galvanometer to be employed both for the resistance and leakage circuits ; but it was found to be more convenient to use two galvanometers , IMPARTED TO A VACUUM BY HOT CONDUCTORS .
S2 .
bility of the Current .
As some unsteadiness had been observed in the galvanometer readings for the leak during the earlier observations , a series of experiments was made in order to examine if the current from the wire varied when the conditions were kept as steady as possible .
When the current was passed through the wire the tube became hot and gas was given off from the walls and from the hot wire , so that it was impossible to keep the pressure absolutely constant .
However , by continuously pumping out the gas the pressure was kept practically constant , the limits of variation being very small .
A constant current was run through the wire so that its temperature and resistance were invariable except in so far as they depended on the pressure of the gas in the apparatus .
Since the rate of escape of heat from the wire is determined largely by the gas pressure , the temperature of the wire is a function of the pressure .
In fact , the galvanometer spot was a far more sensitive indicator of the pressure than the McLeod gauge .
By carefully watching the galvanometer and pumping accordingly , the variations of both pressure and temperature were kept very small indeed .
Under these conditions it was hoped that the rate of leak from the wire with a constant voltage would remain approximately constant .
It was found , however , that it varied in the most haphazard manner , oscillating irregularly between the limits of and ampere .
The current did not become any steadier with continuous heating .
Readings taken eyery three or four minutes for the space of three hours showed the same continuous irregular periodicity .
The irregularities were quite independent of the potential that was or had been applied to the wire , and also seemed to have no relation to the rate at which gas was given off .
There was no measurable falling off with time .
It ought , perhaps , to be mentioned that the tube used for this experiment seemed far more variable than those used for the temperature experiments , though they were never examined systematically .
The platinum wire used for this tube was the purest obtainable .
These results are taken to indicate that the negative ionisation depends to a great extent on small changes in the condition of the surface of the hot wire .
We should expect this to be the case on the view that the phenomena are due to the escape of corpuscles from the metal , since an alteration of 14 per cent. in the work done by an ion in going through the surface would multiply the current by 100 .
Further experiments showed that fairly steady readings were obtained if the heating current was stopped and the tube allowed , as it were , to recover itself between each observation .
The initial value of the current was almost constant , it then began to decrease slowly and afterwards varied in the irregular manner described above .
The following readings taken with constant voltage , temperature , and pressure at the times stated , indicate the sort of agreement is observed : \mdash ; MR. O. W. RICHARDSON ON THE ELECTRICAL CONDUCTIVITy This mode of observation is practically that followed in S5 , where the current is obseryed immediately after the temperature of the wire has been increased by a given amount . .
Experiments with Alte Currents .
A mode of observation which is especially well calculated to show the relation between the positive and negative ionisation produced by hot platinum in a yacuum is to heat the wire by putting it on a 200-volt alternating circuit and to observe the current to the cylindrical electrode .
The ions of both signs are alternately driven away from and attracted to the hot wire owing to the alternating field between the wire and the cylinder .
The cylinder is connected to one quadrant of an electrometer , the other quadrant being put to earth .
The direction of the current , which is indicated by the direction in which the spot of the electrometer moves , is determined by the sign of the ions which reach the cylinder in greatest quantity under the alternating electromotive force .
At low temperatures all the ions produced by the wire are positive , so that the current is necessarily in the positive direction .
At higher temperatures negative ions are also produced in gradually increasing.quantity , so that at one temperature the same number of positive and negative ions reach the cylinder in a given time .
At this temperature , which may be called the transition temperature , there is no current from the hot wire to the surrounding electrode under the given alternating field .
At still higher temperatures , owing to the rapid rate at which the negative ionisation increases with the temperature , the current is always negative .
In these experiments the temperature of the wire was not determined , but a rough idea of it can be obtained from the resistance .
This was determined with the apparatus indicated in fig. 2 , except that the galvanometer was replaced by a telephone .
The following table gives corresponding values of the leak and the resistance of A VACUUM BY HOT CONDUCTORS .
It will be seen these experiments and those to be described later that the negative ionisation increases very rapidly with the temperature , and enormous compared with the positive .
The transition temperature for platinum at low pressures is about 90 C. S4 .
The Relation between the Current and the Electromottve Force .
As the ultimate object of these experiments was to measure the ation current from the wire , it was thought advisable to investigate the relation between the current and the potential applied .
A large number of current-E.M.F .
curves for hot platinum have been given by Professor As , however , my apparatus , though similar , was not quite the same as , and the currents employed were much greater than , the case investigated by Professor McCnELLAND , it was considered necessary to make new experiments on the subject .
As the absolute value of the current was continually varying in the way viously described , the current was valley referred to its value with a given potential on the wire .
This " " standard\ldquo ; potential was yolts .
The current with 41 volts on the wire was measured both before and after taking a reading with any assigned potential : the ratio of this reading to the mean of the readings with 41 volts was taken to be what the ratio of the current under the giyen voltage to the current at 41 volts would have been if the state of the tube had remained constant .
In this way the yariability of the hot wire could be satisfactorily eliminated .
In these experiments the value of the saturation current was about ampere , and was probably about ten thousand times as big as the current used by Professor McCLELLAND .
In making the observations readings of the curTent were taken for every 6 or 7 volts up to 80 , and afterwards at intervals of 40 volts up to 400 .
The numbers so obtained are plotted in the following curve ( fig. 3 ) , the value of the current with volts on the is fixed arbitrarily as unity .
The voltages refer to the positive end of the wire .
The current rises to about its final value with ten volts on the wire but 'Camb .
Phil. Proc vol. 11 , p. 296 .
IMPARTED TO A BY HOT CONDUCTORS .
5 ] 8 making the observations , in order to reduce the evolution of gas as far as possible .
When the tubes were heated at first large quantities of gas came off , but after a time further heating and pumping did not seem to much reduction in the rate of evolution of gas .
Despite constant pumping the pressure always rose slightly in the McLeod gauge ; the increase was , of course , more marked the higher the temperature of the wire .
To obtain the temperature from the resistance measurements use was made of the determinations of the meltin points of potassium and sodium sulphates of Messrs. HEYCOCK and NEVILLE .
The wire was set up in air and its resistance determined , first at the ordinary temperature and afterwards when the smallest possible grain of potassium sulphate placed on it just melted .
In this way the resistance for two temperatures differing by about 100 was obtained , and the temperature corresponding to any other resistance reading could be got by interpolation from the curves given by ofessor CALLENDAR .
To test the method the melting point of sodium sulphate was determined and no determination was more than from the true value This agreement was held to quite good enough for the purpose .
The temperature as found from the resistance in this way is the avel.age temperature of the wire whereas what is required is the temperature at the surface .
A calculation showed , however , that the temperature at the centre of the hot only differed that at the circumference by C. , a quantity which is compared with the experimental error .
In the expel.iments on platinum a potential of 120 volts was main on the wire ; this was more than enough to saturate the current at practically all the which occurred .
It was found afterwards that at pressures greater than 09 millim. the was not saturated by this potential but the deviation the saturation value due to this cause is smaller for the observations taken than the error due to unavoidable irregularities .
The values of the pressures are given in the tables for comparison , Two numbers are inserted in each case in the resistance column , the resistances as determined immediately before and after the value of the saturation current was read .
The difference between the two numbers is a measure of the rate at which the temperature of the wire was changing and , therefore , of the rate at which the pressure of the gas in the tube was increasing .
The following table the results which were obtained at temperatures below 1450o C. 'Jour .
Chem. oc ] .
, p. 160 .
'Phil .
) vol. 48 , p. 519 .
VOL. .
3 O. W. RICHARDSON ON THE ELECTRICAL CONDUCTIVITy being temperatures in C. In the first curve , starting from the left of the diagram , each unit of the ordinate represents ampere ; in each succeeding curve as we pass to the right the value of the ordinate is successively multiplied by ten , so that in the last each unit is equal to ampere .
To obtain the saturation current per ) the values on the curve have to be multiplied by 2.5 .
The curves show that the negative ionisation increases very rapidly with the ture of the wire ( in fact the saturation current varies roughly as the 70th power of the absolute temperature ) .
It will be seen that the current never vanishes absolutely , but only in an nptotic manner , so that it should be observable at any temperatul.e provided sensitive enough instruments are employed .
As a matter of fact at low temperatures it would of course be masked by other effects , which become large by comparison .
The curves seem also to tend continuously to an infinite value of the saturation current ; but the theory indicates that at higher temperatures the current would increase much more slowly with the temperature .
This falling off of the rate of increase has not yet been observed with any of the conductors which have been examined .
We are now in a position to apply formula ( 7 ) to the reduction of the experimental results .
For the sake of convenience we may write for the number of corpuscles shot off from unit area of the metal per second where and .
The saturation current is here to be measured in electrostatic units .
In order to test the formula we may write the above equation in form : If we put , for convenience , \mdash ; and we may write our equation so that plotting the values of against those of should give a straight line .
In the accompanying graph the ordinates are the values of , the cissae being .
The curve got is very approximately indeed a straight line ; though any variation from strict rectilinearity might be explained by the variation with temperature of the coefficient , that is , if our theory is correct , of the number of corpuscles per cubic centimetre of platinum .
We may therefore say with certainty that the main features of the phenomenoll are to be represented by a formula of the type Interesting conclusions are also to be drawn from the actual values of the constants IMPARTED TO A yACUUM By HOT CONDUCTORS .
themselves .
the constant A we the number of free corpuscles in a cubic centimetre of solid platinum , since we have the relation A. A is obtained by putting values of and in the equation \mdash ; At ( absolute ) this gives .
The various constants in the logarithmic equation come from the area of the wire , which was sq .
centim. , and the value of the charge on an ion , which was taken to be electrostatic unit .
The value of being the mass of , and the gas-constant for , one corpuscle ) was found to be .
Putting this in the expression for , we find free ative ions in a cubic centimetre of platinum at 1542o absolute .
An independent value of has been obtained by Mr. PATTERfrom experiments on the change of resistance of platinum in a magnetic field .
This when calculated by the method given Professor yields The agreement of the value found above with this is really very good , when one considers the numerous sources of error to which the measurements are liable , and that an error of 7 per cent. in the absolute temperature , among other things , would multiply the value of by ten .
It was thought that possibly some regular change in the value of with the temperature might be found if for different temperatures .
It was found , however , that oscillated in an irregular manner between and , so that the experiments yielded no any detectable variation of .
This method of obtaining is extremely inaccurate , so that the agreement between the above numbers is really better than would be expected .
The ification of the constant which occurs in the exponential factor is equally important , since is the work done by an ion in passing through the surface layer .
We obtain from the equation where and are corresponding currents and absolute temperatures .
'Phil .
Mag. ' ( 6 ) , vol. 3 , p. 643 .
J. J. , 'Rapports presents au Congr de Physique , ' vol. 3 , p. 138 , Paris , 1900 .
MR. O. W. RICHARDSON ON THE LECTRICAL CONDUCTIVITy Substituting the values of and for espectively , we get the average yalue of from 1378o to absolute as .
If we assume that all the work done by the corpuscles in passing through the surface is electrical , we can calculate from this the discontinuity in the potential .
Since is equal to have , where is the charge on an ion and S is the discontinuity in the potential at the surface of the metal .
From this we obtain electrostatic unit yolts .
The further discussion of these results will be postponed until the experiments on sodium and carbon have been considered .
( See additional Note at end of this paper .
) II .
Exp eriments with Carbon .
S 1 .
Description of In order to detect and examine the negative from carbon , the hot wire previously employed was replaced by a filament from an ordinary incandescent lamp .
The thick filaments from small 8 or 12-yolt lamps were found to be most suitable .
In form of apparatus which was used to investigate the relation between the negative leak and the resistance of the carbon ( see fig. 6 ) the filament was allowed to remain inside the lamp .
The lamps were opened up by snipping off the glass point at the top with a pair of pliers .
The wide tube A was then fixed on by drawing it out at the end which was to be joined and blowing the junction out until it was wide enough to allow the aluminium electrode to be introduced .
This process required some care , as the lamps are liable to crack when hot .
It was found that air leaks due to small cracks in the part of the lamp which is covered with plaster of Paris could be effectually stopped by embedding the whole lamp in melted paraffin wax .
The tube , into which the electrode was fixed with sealing-wax , was joined to a bulb , which was somewhat wider than , into which it was inserted , the joints being made air-tight by means of sealingwax .
The side tube led to pump and McLeod gauge .
The filament could be charged either positively or negatively , and the leak from it to the electrode was measured in exactly the same way as has already been described in the case of platinum .
This form of apparatus was found to be quite satisfactory for investigating the connection between the current from the carbon on the one hand and the electromotive force , the resistance of the filament , and the replaced by the tube shown in fig. 6 .
In commencing an experiment the apparatus was exhausted to 001 millim. , so that the gas which was afterwards in the tube was given off from the hot filament and the walls of the tube .
The observations were genel.ally takell so as to keep the tube as cool as possible , but letting it get hot enough ) to a millimetre could be the McLeod gauge , even with constant pumping .
The first riments were determine the in which the leak varied with the applied electromotive force , other collditions being , so .
as possible , kept constant .
S2 .
Relation between and the E. In all cases there was no current which would show a deflection in the galvanometers used when the filament was charged ) ositively .
The positive leak from hot wires in a vacuum , though large when by an electrometer , is always negligible coml ) ared w the rents m in these experiments .
Some of the subjoined current E.M.F. curves were obtained by using the apparatus in fig. 6 , others by using that in fig. 7 .
As we should exlJect the curves to vary considerably with the and position of the electrodes , the apparatus from which the curves were obtained will be definitely specified in each case .
As an abbreviation for the " " apparatus shown in fig. 6 " " we shall write " " paratus 6 and so on .
The relation between the curl.ent and the electromotive force depends largely on the of the gras in the apparatus .
It may also depend on the value of the maximum current which can be obtained , i.e. , on the temperature of the wire .
At very low ( below , say , 02 millim. ) the current rises very rapidly with the E.M.F. till it reaches a certain value , after which it becomes practically independent of the E.M.F. This " " saturation current\ldquo ; generally increased slightly with the electromotive force , the increase being attributable to the extra ions produced by collisions with the gas molecules .
The following curve , given by apparatus 7 , shows the phenomenon of saturation very clearly .
The flow of the heating current was accompanied by a P.D. of volts between the two ends of the filament , so that it is tant to state which end of the filament the voltages refer to .
There was found to be no current when the negative end of the filament was earthed , the whole of the filament being then positive to the surrounding earthed electrode .
The values are means of a considerable number of servations .
This curve ( fig. 8 ) is for a pressure of 003 millim. ; the voltage given is that of the negative end of the filament .
From the preceding curve it will be seen that the current was practically saturated by a potential of about 15 volts .
With higher of gas in apparatus , the IMPARTEI ) TO A yACUUM By HOT CONDUCTORS .
( 1 ) The resistance of the filaments ; ( 2 ) The currents used to heat the filaments ; and ( 3 ) The temperature of the carbon surface , respectively .
In what follows a section will be devoted to each of the above headings .
S3 .
The Relation between the Saturatvon Current the of the This was inyestigated in the same manner and with the same apparatus ( fig. 2 ) as in the case of platinum .
In all cases the apparatus which has already been described and is shown in fig. 6 was employed .
The thick German silver resistance ( fig. 2 ) , which served as an intermediate standard , had now a resistance of ohms .
The smaller currents were measured with the sensitive Thomson galvanometer .
For the larger currents a D'Arsonval galvanometer , which gave a deflection of 1 millim. for a current of ampere , was employed , owing to its greater convenience .
As the resistance of the filament decreased slightly during the observations , a reading was taken both before and after each observation of the leak , the mean of the two readings bein taken as the value of the resistance which corresponded to the reading for the current .
readings were taken over a range of saturation current extending from to ampere per sq .
centimetre of surface .
The corresponding range of the value of the ratio of the resistance of the filament to its resistance at C. at the commencement of the experiment was from to 567 .
In other words , while the resistance of the filament only alters in the ratio of 610 to 567 , the negative leak has become twenty thousand times as big as it was at first .
It is evident , therefore , that , as in the case of platinum , the number of negative ions produced at the increases with enol.mous rapidity as the temperature rises .
It will be shown later that , over a much greater range of temperature than this , there is no perceptible falling off in the rate at which the current increases .
The corresponding values which were obtained for the saturation current and the resistance of the carbon filament are given in the following table .
The resistances are expressed as fractions of the resistance which the filament possessed at C. before it was heated .
These numbers when plotted against one another on squared paper yield curves very like the current temperature curves for platinum ( fig. 4 ) .
They have not , however , been inserted , since they are much the same as the current-temperature curves for trbon , which have been plotted from the same observations .
Just as in the former experiments , the current was never found to be a function of the electromotive force alone , so also here the cooling curves neyer exactly coincided with those obtained as the temperature of the filament was raised .
This was partly due to the permanent change in the resistance of carbon produced by heatin It was attempted to correct for this by taking , instead of the ratio of the resistance at moment to the original at C. before commencing the experiment , the ratio to the resistance which the filament would possess if at a temperature of C. at that moment .
The permanent change in the resistance was assumed to be proportional to the rate at which the resistance changed during an experiment , the conditions being kept , as far as possible , constant .
In this way it was possible to obtain by extrapolation the resistance which the filament would possess if allowed to cool down to C. at any stage during the experiments .
That this process brings the two curves more nearly into coincidence will be seen at once on comparing the numbers in columns I. , VI .
, and VII .
of the following table .
this case the potential on the filament was volts , and the heating current was run at constant voltage .
The results of these corrections are shown in the accompanying table :\mdash ; MR. O. W. RICHARDSON ON rfHE ELECTRICAL CONDUCTIyITy It will be noticed that there is a great similarity between fig. 14 and the curves connecting the current with the esistance and temperature respectively .
This is merely due to the rate of variation of the saturation current being so rapid that the differences in the alteration of resistance , temperature , and heatin current becomes insignificant in comparisQn .
Since in the last experiments the big currents were always accompanied by a high pressure of gas in the apparatus , it might be thought that part of the increased current was due to the gas present .
To investigate this point a series of experiments was made with the apparatus shown in fig. 7 , so that the gas pressure could be kept down to a very low value .
By heating the filament for a short time only and taking the observations very quickly , it was found that the temperature of the bulb and electrode could be preyented from rising perceptibly .
Under these conditions it was found that the amount of gas given off was greatly diminished , the highest pressure recorded during the observations being 006 millim. A reading of the McLeod gauge was taken between each reading of the galvanometer deflection for the saturation current .
The potential on the filament was volts , this being more than enough to saturate the current ( cf. fig. 8 ) .
The numbers which were obtained are given in the accompanying table : The greatest value of the saturation current is not given in the above table , since the corresponding reading of the ammeter was not taken .
This enabled the reading to be taken much more quickly , so that the pressure only changed from 0022 to 0025 millim. The corresponding saturation current was 04 ampere ; in other words , a square centimetre of surface would have given a current of 28 ampere across a vacuum at millim. pressure .
In the preceding series of experiments the highest possible value of the temperature had not been reached , so that further experiments were instituted to determine the maximum current which could be obtained from a square centimetre of a carbon ament when the temperature was pushed to the highest limit , i.e. , just before the IMPARTED TO A VACUUM BY HOT CONDUCTORS .
filament melted .
The arrangement of the apparatus was altered somewhat , the leak being measured by a Weston ammeter instead of a galvanometer as before ; in other respects the arrangement was unchanged .
The pressure was kept very low and a potential of about volts was maintained on the filament , the surrounding cylinder bein earthed .
With this apparatus it was found possible to maintain an actual current of 4 ampere ( corresponding to 2 AMPE per square centimetre of filament surface ) at a pressure of less than millim. The current could not be made to surpass this value since the filament melted on raising it to a slightly higher temperature .
The fact that such large currents can be produced at such pressures has an important bearing on the theory of the mechanism by which the corpuscles produced , hich will be considered later .
S 5 .
The Relation the Current the Ternperature .
From the experiments on the variation of the saturation current with the resistance we have been able to give numbers which indicate , roughly at any rate , the way in which the former depends on the temperature .
It was thought that a more reliable estimate might be obtained if the temperature of the filament were determined by means of a thermal junction of platinum and iridio-platinum .
With this object the following experiments were made : The tube employed was that shown in fig. 7 , and already described , The filament in this tube was in the fornu of a simple and had the following linear dimensions : length centim. , diameter centim. , and total of surface sq .
centim. For these experiments the apparatus shown in fig. 2 had to be altered , the portions below being entirely reconstituted .
The apparatus used for measu the saturation current was unchanged , the only alterations being made in the pol.tion used to measure the temperature .
The thermocouple circuit was completed by taking the lead ( fig. 7 ) to the cold junction , which wa.s placed in a test-tube immersed in water at C. ; the other wire the cold junction passed through a resistance box to a D'Arsonval galvanometer , and thence through ( fig. 7 ) to the hot junction .
The adjustable resistance ( fig. 2 ) still seryed to regulate the current which was used to heat the filament .
In order to standardise the ttlermocouple the melting-point of potassium pbate was again taken as the fixed point .
Ajunction of the same wire as that used during the experiments was fixed on to a stout , which was clamped horizontally in the hottest part of a Bunsen Bunsen was arranged to burn vigorously with a bright green inner cone and was carefully protected from draug.hts .
Very small portions of the salt were then placed on the stout on the side of the flame opposite to the thermocouple , and mattel'S were so arranged that when the salt just melted it was exactly the same distance from the edge of the flame on the one side as the thermocouple on the other .
he reading of the galyanometer was then MR. O. W. RICHARDSON ON THE LECTRlCA L CONDUCTIVITY taken to correspond to the melting-point of the salt .
The method was then tested by a small portion of salt on the thermocouple itself and observing when it to melt .
This was found to agree with the previous observations .
The greatest difference between the observations taken was less than 3 per cent. further test was supplied by determining the melting-point of sodium sulphate ; the value found was within of given by Messrs. HEYCOCK and NEVILLE .
This agreement was considered to be quite good enough for the purpose in hand .
The couple was finally found to give an electromotive force of millivolts when its junctions were at 1067o C. and C. respectively .
The platinum temperatures given by the galvanometer readmgs have been corrected to bhe air thermometer scale by means of the curves given by Professor CALLENDAR .
The thermocouple method possesses one great advantage over the resistance method of determining the temperature of hot wires , in that the observations can be taken much more quickly , and so the wire has to be heated for a much shorter time .
In this way the apparatus need never get hot , and far less gas is given off , so tha .
the readings generally are much steadier .
The accompanying table represents a series of observations with this apparatus .
The pressure was always less than of a millimetre of mercury , while the potential on the wire was volts , this being more than enough to saturate the current .
The platinum temperatures are given under the column headed Pt , the numbers under are the temperatures ( degrees Centigrade ) reduced to the air thermometer scale .
It will be noticed here again that the current for a given temperature is smaller as the temperature is being increased than when it falling .
'Chem .
Soc. Journal , ' vol. 67 , p. 160 .
'Phil .
Mag vol. 48 , p. 519 .
IMPARTED TO A VACUUM BY HOT CONDUCTORS .
The following table represents a series of observations taken with the thermocouple apparatus at somewhat higher tempera)tures .
The potential on the filament was here volts .
The temperatures given by the thermocouple method are on the average about lower than those obtained from the resistance for the same current .
The temperature registered by the couple would be lower than that of the more remote parts of the filament for several reasons , the chief one being the conduction of heat away locally by the leads of the thermocouple itself .
It is difficult to say whether we should expect this difference to amount to 12 C. We are now in a position to test whether the experimental results are in agreement with the theoretical formula for the saturation current , .
:\mdash ; using the notation employed before .
If we take , as in the case of platinum , and , the above equation reduces , as before , to the straight line The following curve ( fig. 15 ) has been plotted in this manner the numbers given in the table on p. 525 .
The ordinates are values of the abscissae are values of MR. O. W. RICHARDSON ON THE ELECTRICAL CONDUCTIyITY The numbers on pp. 530 and 531 yield the curves in fig. 16 .
The straight line on the right is drawn from the observations on p. 530 , and corresponds to lower temperatures than the other .
The experimental points for this curve are denoted by .
The other curve from the obseryations on p. 531 refers to somewhat higher temperatures .
The experimental points for this curve are indicated thus : All these three curves are fairly close approximations to a straight line ; it is therefore quite evident that the observations are represented very closely by assigning to the saturation current a formula of the type When we come to the actual values of the constants in the above formula , the agreement with the simple theory is not so good as in the case of platinum , though possibly this is partly due to the greater difficulty of the periments .
The curves in figs. 15 and 16 give for the value of , and ] respectively .
In order that the differences of the values of should be proportional to the contact E.M.F. between carbon , platinum , and other metals , for carbon should be since its value for platinum is .
The difference between this and the above numbers does not appear to be very great , but the effect of a small error in is enormous when we come to calculate from it the value of , the number of corpuscles in a cub. centim. of carbon .
If we take as the best value of , and at 1515o absolute as being the mean of the two series of temperature measurements , we find A is of the order and is of the order .
Now , Mr. PATTERSON* finds that at ordinary temperatures .
The effect of temperature on the resistance of carbon indicates that the concentration of the corpuscles would be at least ten times as great at 1000o as at C. , so that we should expect to find of the order .
As a matter of fact , if we take instead of , we find instead of Reasons which might make this method of determining give values which are too large , will be considered at some length after the experiments on sodium have been described .
III .
ents with S1 .
Nature of Sodium was selected as the next metal to be investigated on account of its strong electropositive character .
Since this implies a great attraction for positive electricity , we should expect its power of retaining the negative corpuscles to be much smaller than that of the conductors hitherto examined .
If the foregoing theory is correct the corl)uscles ought to escape from the alkali metals at a much lower temperature than ' Phil. Mag 6 , II1 .
, 655 , IMPARTED TO A yACUUM BY HOT CONDUCTORS .
from metals which are low down in the volta series .
In fact , assuming ( 1 ) that the difference in the discontinuity of potential at a platinum vacuum and sodium vacuum surface is equal to the contact difference of potential for sodium and platinum 'taken roughly to be equal to two volts ) , ( 2 ) that the value of the discontinuity ( volts ) previously obtained for platinum is correct , ( 3 ) that the concentration of the corpuscles for sodium is of the same order of magnitude as for copper , and ( 4 ) the correctness of the present theory , a preliminary calculation showed that currents of the order of some ampere per square centimetre ought to be obtained at as low a temperature as 50 C. The problem we have to face in the case of sodium is not quite the same as in the case of llon-yolatile substances such as carbon and platinum .
For in this case the metal has an appreciable vapour pressure at the temperature at which the experi- ments are carried out , and part of the conductivity present is doubtless due to the spontaneous ionisation of the metal vapour .
A second inconvenience , which is more of a practical nature , is caused by the distillation of the metal from the hotter to the colder parts of the tube , causing the state of the latter to continually vary .
For the same reason some of the sodium condenses on the electrode which is supposed to be free from it , so that both electrodes emit negative ions .
We have seen that in the case of platinum and carbon no current was obtained when the hot conductor was positively charged ; in other words , conductivity was perfectly .
In the case of sodium , owing to the spontaneous ionisation of the vapour and the condensation of the metal on the inserted electrode , we should expect to get a current in both directions .
In the following experiments the first effect must have been small , owing to the low vapour-pressure at the temperatures employed , while the effect of the second was made small by using an electrode with a very small superficial area ( a thin platinum wire ) .
It will be seen that in every case the current when the sodium surface was negative was more than twenty times its value when the surface was positive .
The apparatus which was used to detect and measure the ative leak sodium will now be described .
S2 .
Description of After a great number of trials of various forms of glass apparatus , all of which came to an untimely end owing to the joints not being able to stand the continued heating or otherwise , the metal apparatus shown diagrammatically in fig. 17 was set up .
The weldless steel tube ABDC was 76 centims. long and .
in diameter , and was kept at zero potential by means of the earth wire shown .
The straight platinum wire was insulated with sealing-wax at each end and could be charged positively or negatively to any desired potential .
It formed the electrode mentioned above .
The whole tube was placed a small combustion furnace , by means of which the central portions could be heated to any desired temperature .
The temperature was determined by means of a copper-nickel thermocouple , , attached to a hollow semicircular cylinder of brass , , placed at the middle point of the line .
The brass piece was cut from a tube which before the operation fitted easily into ABCD , so that when the copper and nickel wires were tied round it , it fitted quite tightly .
The spirals were made of composition tubing wound tight ound ABDC ; they served to keep the ends of the tube cool and thus prevent the sealingwax joints softening .
They were fed with cold water at and emptied at G. The side-tube from led to the pump and McLeod gauge .
The manner in which the wires and were fixed in at the ends is shown more clearly in the enlarged diagram on the left of the figure .
The shaded parts represent the distribution of the sealing-wax which was used to make the joints .
The ends of the tube ABDC fitted into an annular depression on the brass plate aa .
The platinum wire was soldered at each end on to a stout copper wire , which fitted fairly tight in a glass tube passing through the brass tube .
The whole was fixed in air-tight by means of sealing-wax .
One of the leads from the thermocouple was fixed in exactly the same way into the brass tube The sodium was originally placed in the form of small cubes on and around , and it was considered that after heatin for a short time in a vacuum a fairly uniform distribution of sodium over the central portions of the steel tube would be obtained .
This was certainly what happened in the case of the glass apparatus which had been tried previously and in which the effect could be observed .
The leak from the hot sodium to the platinum wire electrode was then measured , according to its magnitude , either by an electrometer or by a galvanometer .
At the lower temperatures where the quadrant electrometer was employed , one of the quadrants was connected to the case of the instrument which was insulated .
The other quadrant was connected with a standard condenser and , by means of a wire passing axially on sealing-wax supports along a brass tube , with the electrode .
The outside of the brass shielding-cylinder was connected with the case of the electro- meter .
In making an experiment the whole of the electrometer system was charged to a given potential , and the time required for the spot to move over a given number of scale divisions was noted .
This measured the current from the insulated electrode TO A VACUUM By HOT CONDUCTORS .
5.3.5 to the surrounding earthed tube ABDC .
By altering the capacity of the condenser a suitable rate of movement of the spot of the electrometer could be obtained each time .
In using the galvanometer the arrangement was practically the same as that employed before .
One end of the battery was put to earth while the other was connected through the galvanometer to .
The battery was capable of ying any number of volts up to 420 , the potential being measul.ed by the Weston voltmeter used previously .
A D'Arsonval galvanometer giving 1 millim. deflection for ampere , and having a resistance of 500 ohms , was used .
With the exception of the change in the materials of the couple the thermoelectric circuit was exactly the as that employed in the expel.iments on carbon .
To reduce the alvanometer readings to temperatures use was made of the recent observations of Mr. E. P. HABRISON .
* Only one fixed point was determined , that of the boiling-point of sulphur .
The electromotive force at that temperature was found to correspond to microyolts per degree over the whole range , a result which agrees very accurately with that given by Mr. HARRISON .
The relation between electromotive force and temperature was not assumed to be linear , but corresponding values for intermediate points were calculated from Mr. HARRISON'S curves .
From these figures a curve was plotted which gave temperatures in terms of galvanometer readings directly .
The galvanometer employed gave 1 millim. deflection for ampere , and the total resistance of galvanometer and lectric circuit with no resistance out of the box was ohms .
In the various forms of glass apparatus previously tried it was found that considerable currents were obtained at ordinary temperatures when the sodium was charged negatively .
This was ultimately found to be due to the photoelectric effect produced by the light presen in the room , since it disappeared when the experiments were ma.de in the dark .
The steel tube finally used in the experiments had the great adyantage that it could easily be mffie absolutely light-tight .
In to make sure that no light reached the sodium , the glass tubes through which the wires were let in at each end of the steel t , ube were painted over with black enamel .
The leak was then tested and found to be small and the same in both directions ; so that it was all due to imperfections in the insulation .
S 3 .
The Relation between the Current and the Force .
After testing the insulation and pumping down the apparatus experiments were first made to see how the current varied with the direction of the electromotive force .
The first measurements showed that , at a ) temperature of about 30 C. , the current when the wire was at a potential of volts was times its value when the wire was charged to volts .
The value of the current when the wire was positive was ampere .
Later experiments showed , however , that the ' Phil. , vol. 3 , p. 177 . .
O. W. RICHARDSON ON ELECTRICAL CONDUCTIVITy positive current invariably about 30 times as big as the negative .
This was probably due to sodium having conden sed on the wire electrode .
For , although the wire had a surface per unit-length of less than one-hundredth that of the steel tube , the surface ionisation ) be far easier to saturate ; so that we should expect the currents in the two directions un der a given voltage to have a ratio considerably less than t00 to 1 .
The above high value of the ratio obtained initially would correspond to the stage when no sodium had condensed on the wire .
In making these experiments the apparatus was first pumped down to a pressure of about millim. , but it was found that on heating the steel tube a considerable amount of gas was given off .
At first the amount of gas evolved was so great that the in the atus , the volume of which very considerable , rose to several centimetres of mercury .
This evolution of gas was noticed in every case when sodium was heated , but by continued heating it usually became very small .
In this particular instance , even after heating for several days , on pumping the apparatus out and heating again it was found that the pressure rapidly rose to about 5 millims. It was thought that the gases from the furnace might perhaps diffuse through the steel tube .
To prevent this the latter was covered with a layer of soluble glass , which was carefully dried on ; this seemed to have the desired effect , for it was found that vards there was no difliculty in keeping the pressure below a even when the tube was heated to 45 C. Experiments were next made to investigate the way in which the current varied with the potential when the wire was charged positively .
It was found that the current E.M.F. curves were markedly different from those previously obtained with carbon and platinum .
The current was small at first and increased much more rapidly with the voltage than if the two were proportional .
In fact , the general shape of the cul.rent E.M.F. curves was much like that of the curves for current and temperature obtained with carbon and platinum .
There was no indication of saturation at any potential .
These differences are to be attributed to the difference in the experimental conditions and especially in the shape of the electrodes .
In the case of sodium we have a large ionisation produced at the inner surface of a wide tube , and it is a well known fact that it is difficult to saturate the current to a wire inside the tube in such a case , owing to the weakness of the electric field near the surface .
The accompanying curve ( fig. 18 ) gives the relation between the current and the TO A VACUUM By HOT CONDUCTORS .
electromotive force for voltages in the wire electrode between and .
The pressure was about 5 millims. The sudden increase in the current between 4 and 8 volts was obtained every time and did not seem be due to experimental error .
The current E.M.F. curve from 40 to 240 yolts is very similar to that in fig. 18 , except that it approximates very closely to a straight line between 160 and 240 volts .
It is given in fig. 19 .
When the voltage was increased above 240 it was found that the current rose rapidly to several thousand times its previous value .
The increased current was quite steady at 320 volts , but at 280 volts it seemed to be in a very unstable state , since all kinds of intermediate readings could be obtained .
Above 320 volts the current increased in a linear manner with the voltage .
The experimental numbers are given below .
These numbers seem to indicate that with potentials greater than 240 volts an ordinary vacuum discharge took place at some point or points in the tube ; in the following experiments care was therefore taken never to use potentials greater than 80 volts S4 .
the under given Voltage the In the case of sodium , owing to the fact that the current could not be saturated , its value under given force was measured at different temperatures .
This comes to practically the same thing as measuring the saturation current , since we should expect , ceteris paribus , the current with a given electromotive force always to be prcportional to the number of ions liberated at the metal surface .
In order to be sure of not getting a discharge , a potential of about 80 volts between the wire , and the cylinder was always employed .
The following table represents a series of observations of current and temperature ranging from 21 C. to 42 C. It ill be seen that the range of current is from to ampere ; in other words , raising the temperature of the metal from to increases the current to ten million times its original value .
The currents below ampere were measured with the electrometer .
In this series of experiments very low values of the currents were not measured ; in a later series the current was taken nearly VOL. CCI.\mdash ; A. 3 IMPARTED A ACUUM BY HOT S. .539 We now come to the application of the theoretical formula to the reduction of these results .
This equation may be written as before where being the absolute temperature .
To test the theory , values of have been plotted yainst values of in the following curve ( fig. 21 ) .
It will be seen that all the points ffill very nearly on a straight line except the first two .
They are all , however , fairly accurately represented by the dotted curve shown .
As the two lowest points correspond to a low temperature , it is possible that some other effect is coming in here which would account for their deviation from rectilinearity .
Tn calculating the value of we may either confine our attention to the straight part of curve 4 , and neglect the two first observations , or we may take the average ovel the whole range of the ments .
The two values differ by about 24 per cent. ; if we take the mean we find .
This gives , for the discontinuity of potential at the surface , the value volts , and would therefore give volt as the difference of its values for sodium and platinum .
The above value of the difference , which is approximately equal to the contact electromotive force for sodium and platinum , forms a strong confirmation of the theory ; but when we come to calculate , the number of free corpuscles in a cubic centimetre of sodium , the agreement is not so good .
In fact , we find from the experimental results that for ; whence , putting in an estimated value of the area of the sodium surface , we get about .
The value of has not been determined for sodium by any other method , but we should expect it to be not greatly different from that for copper , which is given by Mr. PATTERSON as .
The value given by this method is thus far too great , for it is hard to imagine that the corpuscles can have a pressure of ten million atmospheres .
I believe the discrepancy here is greater than can be explained by errors of experiment , although that was possible in the case of the high values found for carbon .
A further series of current measurements was made , using the quadrant electrometer , in order to investigate the leak at somewhat lower temperatures .
It was also thought desirable to measure the currents under a given voltage in each of the two possible directions and to see if there was any relation them .
The experiments also served to test whether the current temperature curve , obtained when the tube was followed the same path as that obtained with rising temperature .
It was scarcely to be expected that curves would coincide , even approximately , owing to the continual distillation of the sodium from the hotter parts of the tube to the cooler .
The results of the experiments are given in the accompanying table .
In making the observations readings were generally taken with the wire electrode alternately positive and , a potential of 84 volts being used .
In taking each reading the capacity was adjusted so as to give a convenient rate of movement of the electrometer spot .
It was attempted to take corresponding positive and negative readings at as near the same temperature as possible , the gas furnace being adjusted after each pair of readings had been taken .
The pressure in the apparatus varied from 25 to 4 millim. The meaning of these numbers is best expressed graphically .
In fig. 22 the arithm of the current has been plotted against the temperature .
The unit of current is ampere .
Curves 1 and 2 were taken with the wire charged positively , 3 and 4 with the wire charged negatively .
The observations for curves 1 and 3 were made simultaneously with the temperature of the tube rising , whereas the curves 2 and 4 correspond to the second set of observations with the temperature falling .
various marks refer to observational points for the different curves as follows : to No. 1 , to No. 2 , to No. 3 , and to No. 4 .
The value of A is determined from that of and so depends very largely on the value selected for .
An error of 10 per cent. in multiplies A by about 100 , whilst if were determined to one part in three , A would be multiplied by this reason only the order of magnitude of A has been given in the table .
If we can assume that A and are constants independent of the temperature , we obtain the value of , the number of free corpuscles per cubic centimetre of the conductor , at once from the theory by dividing A by .
Treating the values of in this way , we find that the value of for platinum rees satisfactorily with that obtained by Mr. PATTERSON .
On the other hand , the vaiues for carbon and ( 10 ) for sodium are greater than the maximum possible value .
Moreover , the error in each case seems greater than can be accounted for by expelimental uncertainties .
This error is probably due in part to the assumption that A and are constants , whereas it is evident that they must both be functions of the temperature .
It is possible on the preceding theory to say something about the forms of these functions which indicate that they both vary with the temperature .
With regard to the number of corpuscles per cubic centimetre of metal , we suppose they are formed by decomposition of the neutral atoms in much the same way as in any case of chemical dissociation .
If be the number of positive and of negative ions per cubic centimetre ( as a rule ) , being the umber of undissociated atoms per cubic centimetre , and being the value which would possess if there were no dissociation , then : , since the number of re-combinations per second is proportional to , whilst the number of dissociations is proportional to , and these two must be equal in the steady state .
Now VAN HOFF* has shown that for all re-actions of this type the quantity varies with the temperature according to the equation being the heat evolved when two ions re-combine , whence A must be very large , for when , ( A ) must be great compared with We may write in the form whence we see that when , and ; when is large ompared with and We see from the nature of the above function that the value of would decrease 'Lectures on Phys. Chem vol. 1 , p. 141 .
IMPARTED TO A yACUUM By HOT CONDUCTOR with enol.mous rapidity in the neighbourhood of the absolute ; so that , although the resistance of metals decreases steadily with decreasing temperature down to the lowest temperatures yet reached , it is quite possible that it becomes infinite again at the absolute zero .
The fact that the resistance of pure metals is proportional to the absolute tempel.ature over a wide range , together with the high values of which prevail at ordinary , seems to indicate for most metals has practically reached its maximum value , where it varies only slightly with For this reason we are led to the conclusion that the discrepancies of are not due so much to disturbances produced by its temperature-variation ( except , perhaps , in the case of carbon ) as to the fact that the exponential coefficient is a function of .
We have seen that , where is the work done by a corpuscle in escaping from the metal , and is the gas constant for a single .
Now roughly , and for platinum , so that is equal to A second approximation to the value of is obtained when we consider the nature of the forces which retain the corpuscles inside the metal .
These are a sort of integrated effect of the attractions of the positive and negative ions scattered about in the metal near the corpuscle .
The field would thus be much the same as if the corpuscles were surrounded by a perfect spherical conductor of molecular dimensions .
The quantity is therefore of the order as the energy required to remove a corpuscle from such a charged sphere , which is , where is the charge on an ion and is the radius of an atom .
Taking centim. , this gives If this view is correct\mdash ; it hardly seems likely that the above numerical agreement is entirely a coincidence\mdash ; we should expect the value of to decrease as the temperature is raised owing to the greater distance of the atoms apart .
We should therefore expect to decrease in much the same way as the linear dimensions of the metal ease with the temperature .
It is probable , , that can be represented with sufficient accuracy as a function of temperature of the form .
Writing the equation at the beginning of this section in the form we see that the first three terms ( with the possible exception of , which we are not considering ) vary extremely slowly with , if at all , so that we 1nay use as an approximation where .
If now we put , we see that .
So that , as we found in the experiments : is a linear function of , but the constant A which is determined is much larger than it ought to be , owing to part of having become added to it .
As a numerical xample , we may give to the value , which corresponds to a rature coefficient of 00014 per degree , and would change the value of by 20 per cent. in a range of 1400o On calculatin out we find that this small temperature coefficient would leave practically unaltered , but would make the apparent value of one thousand times its true value , whilst doubling the coefficient would square the error in , and so on .
It is therefore evident that the temperature variation of is quite adequate to explain the large values of which have been found .
Moreover , owing to the peculiar nature of the functions , it is impossible to arrive at the true values of by this method .
The value of A found in these experiments are therefore not irreconcilable with the values of given Mr. PATTERSON , but the two values of can be made identical by assigning to a small temperature coefficient .
The coefficients necessary have been calculated , and , together with corresponding orders of magnitude of A and of , are given in the following table .
SodiumCarbonPlatinumConducto..r . . . .
A. Order oemperature coefficientto g From the values of we can calculate the work done by an ion in passing through the surface , and hence the discontinuity of potential between the metal and the surrounding space .
For the case of platinum this has already been done , the value obtained being yolts .
For carbon and sodium , taking into account the temperature coefficients given above , we find for the discontinuity at C. the values and volts respectively .
It will be noticed that these numbers follow the same order as the Volta series , though their differences ( at any rate for carbon and platinum ) are not equal to the corresponding contact electromotive force .
S2 .
The work done by Corpuscle in passing through the Surface Layer .
It has been shown on p. 543 that the value of is of the same order of magnitude as , where is the charge on an ion and is the radius of a molecule ; it is therefore also of the same order of magnitude as the energy set free when two ionoe of opposite sign re-combine , and as the work required to produce an ion by collision .
Theoretical considerations , in conjunction with the experimental results , render it probable that may be represented very approximately as a linear function of the is the absolute temperature .
It is noteworthy that this number .
00022 is practically equal to the coefficient of cubical expansion of sodium ( .
000204 ) .
TO A VACUUM BY HOT 1 ) temperature , whilst the numbers given on show that the coeiJicient of is of the same order of gnitude c coefJicient of lineal expansion of corresponding solid conductol These render it probable tlJat is function of the and distance of the molecules of which the conductol consists .
If we consider in the as the distance apart of the centres of the molecules in the solid state , it will be to the cube root of the atomic volume .
We shonld expect the work done by a corlJuscle in passing the metals to be approxinlately equal to a nt divided by the cube root of atomic .
Up to the present has been determined only for sodiunl , ) and cal.bon , but fortunately these three elements ful.nish a of tomic volume .
As ttter of fact , carbon has the smallest tonlic volulne of elements , whilst that of sodium is only exceeded by the of tomic weight .
In the accompanying values of the atomic volume and the invel.se of its ctll , root in the first two columns .
The third contains the disconliuuily in the potential S , which is ) to ; whilst in last the tios of those in the second .
In the of is some doubt to what the value of the volume should be , density rent values the .
Thus the density is aphite 2 , itnd for diaulond It will be seen the numbers in the last column not quite ; but they only change by about 40 .
cent. , whilst the atomic volume es in { he ratio of 6 to 1 , the atomic weight to 195 .
It seems to conclude that the done by a corpuscle in through the layer is , to a first , il to the cube of atomic volume of the element .
S S. The of Gas on The negative leak heric pressure is always low is ed in both .
In one , lvhe the wire was current of ) of V0L . .
O. ON ICAL CDUCTIVITY let into the ] ) the culre1lt again measured under nospheric temperature .
It was found that there was no detectable leak till a ] ) of highel reached , when one division of the galvanometel alnl ) was obtained with the wire at a potential of volts .
small value of the currents at ) pressure is probably due to the difficulty of ating them .
Tt ) conceivable that the ionisation low pressures might be due to the olecules h the hot wire and becoming ionised thereby .
If we ssulne the ' would COl.respond to each molecule produciug one ion , we calculate its value in any given case .
If we take the numbel of molecules in etre of at C. and 760 millims. to be , then the number hit unit of the wire per second is } oximately , where ( the square root of the mean velocity squal.e ) may be taken as centims. second for .
The number which strike unit area of wire per second at 1 is , which gives a saturation current of Ilits , amperes per square centimetre .
At a this cul'rent would become ampere per square centimetre .
As a matter of iments , a current of alnl ) per entirIletre obtained lnillim . .
This is twenty-five times the ximum value obtained by supposing each molecule to one ion ; so it is ( any considerable ) of the conductivity is due to ions oduced i this way .
Another of considering this question is to calculate the number of ulecnle of inside the cylindrical must collide with the filament second to produce the obseJ.ved current , assuming that each collision sets one .
In the experiments in question cylinder had a volume of about .
centim. , so each molecule would have to pass and between the filament and the cylinder some times each second .
This seems to be an impossible feat for an uncharged molecule .
Both these points of view lead to the conclusion that the corpuscleti not luced by dynamical action between the molecules of the surrounding face of the metal .
In fact , all the experimental results seem to point to that the Ctl.puscles are produced from the metal by a process to tion .
The effect of the ounding g , of impurities in the , nnd of its to be due to alterations in ) of hich ccsponds to latent heat in the theory of evapol.ation .
4 .
ct .
lt be seen results which ined flish i Ldison ) this connecting , ) TO A By IIOT placed the ends of the fihment ndescent 1 ] a to the positive end of the filament .
A current was then obsel.ved which amounted in some cases to several milliamperes , though there was no current the electrode was.joined to the negative terminal .
Evidently the current was by corl)uscles p from the ative portions of the ot carbo to the relatively ositive e , on this vie should pect the current to nish 1 mpari { the electrode ) ative with respect to the filmment .
This observation confirmed and extended Professor using electrodes he aced round rious parts of the filament , that the current only the negative end .
He also found , agreement with the results of the present ) aper , that a filanlent likewise an .
This was in the same direction as , but greater in magnitude than , that by carbon .
Finally , the dison effect found to increase rapidly with the temperature of the filament , which confirms its identification with the phenomena here investigated .
S 5 .
The Energy Emitted .
It is of interest to compare the energy lost by a hot body owing to the emission of corpuscles with the energy given off in form of electro-magnetic radiation .
The recent surements of F. show that the energy radiated in 1 second from 1 .
centim. of the surface of an absolutely black ) at 1o ) solute is ' whilst we lnve seen that the total rate of loss of energy of a conductor owing to the emission of corpuscles at temperature absolute is Since the quantities in this formula have now been ed and platinum , we can calculate at any temperature for these substances .
The first term in brackets represents the part of the energy due to the motion of translation of the emitted corpuscles , and is less than 5 per cent. of the second term at all temperatures at which expel.iments have been carl.ied out .
We may therefore leave it out to a first approximation and calculate only the second term , which is equal to the work done by the corpuscles in passing through the surface layer .
Thi , is obviously equal to [ 5 ] , vol. vol. 6 , O. ON DUCTIVITY S is the discontinuity in the ) , and is turation C in is ( that for ) tinum at ] ] . .
The value is a much larger quantity , ntim .
sec. The largest experimental values of were obtained with carbon .
ince the greatest value of the uration current attained was a]nl)ere lllits per square centimetre and S ) volts , we ) the lost wire fcentinl .
sec. The ondi l to this current not ]he{sured , but was certainly ter than 2000o energy radiated from an absolutely black body at absolute would have been ergs/ centim. sec. We see then all temperatures at which experiments were made the of due to the } ) of the corpuscles is much less than that due to the emission of ordinary omagnetic radiation ; on the other hand , it increases much more rapidly with the temperature , so that , in the case of carbon at any rate , it would become first equal to , and finally great compared with , the netic radiation , at temperatures not much above 2000o C. It must not be forgotten that for this calculation the hot conductor is supposed to be placed in a vacuum and surrounded by an electric field which removes the ions ; otherwise all the ions diffuse back to the metal and is no loss of energy due to this cause .
all these experiments we a long way from the region where an appreciable ctio of the total number of ions which strike the surface of the conductor pass .
This is easily seen if we calculate the value of the saturation current ) on supposition that every corpuscle which hits the surface escapes .
Let 11H .
as a probable maximum for the number of corpusdes in a timetrc of , say , carbon ; then , at 2730o absolute , so that the aturation current would be ] electrostatic ) amperes per square cenfimetre .
As the current which has been yet ined is amperes per centimetre , it is evident that we are still a long way from the limit .
This c , qlculation seems to indicate that the region on the current temperature diagram when the current begins to be proportional to the square root of the absolute temperature is much higher than any temperature which can be reached in ordinary way .
The magnitude of the currents which have been obtained with low voltages , that a ) vacuum bounded by a hot conductor is , at any rate under certain tmlces , an extremely good conductor of electricity .
In fact , it seems probable htcl } cunm is capable of becoming the best conductor that can possibly be .
The conductivity of metals is limited by the shortness of the mean free of ions , whereas the mean free path of a corpuscle in an atmosphere of is probably large .
All that is necessary , therefore , to produce current is to ) ) ) ions quickly at the hot , that is , to raise the By HOT ) ture of the hot conductor to sufficient extent .
) ) ) show nps electrical conductivity is concerned , the boundary of a hot conductor is an indefinite term ; since so many of the corpuscles pass freely outside of the metal it is evident that at ] ) enough reci b of th cnrrent ( wire lllHt ) } ce .
conclnsion , wish to thank Professor .
advice enlent during the course these periments , hidl 1 carried out in the orntory .
[ Notc , added , 1903.\mdash ; Since the present paper was written Mr. H. A. WILHON has made some experiments on the conductivity produced by hot platinum at low pressures , in which he finds that by carefully treating the wire the current can be reduced to about one two hundred thousandth of the value found by the author at the same temperature .
Mr. WTLSON also shows that the current is greatly increased by admitting hydrogen into the apparatus , and concludes that the high values found in this paper are due to hydrogen absorbed by the wire , which is only given off slowly , if at all , by mere heating .
These results are not , however , inconsistent with the view that the effects are due to electrons shot out of the metal .
To obtain the observed facts we have only to suppose that the occlusion of hydrogen diminishes the work which a ) corpuscle to do in escaping from the surface .
Mr. own results in agreement with this theory , for he finds that raising the pressure of hydrogen from to millims. luces the of the work in question in the ratio of 155 to 36 .
It ] be thought that on this view the constant hich determines the number of ions per cub. centim. of platinum should be independent of the pressure of the outside .
The numbers found by Mr. WILSON do not support this supposition , but the numerous practical and theoretical difficulties that little should be attached to the difference .
It is possible that Mr. WILSON'S process of removing from a wirG by oxidation may , as it were , overshoot the mark by leaving an electrical double layer with negatively charged oxygen on the outside .
Such a double layer } ] increase the work for the corpnscles to ont and so the the seryed .

