]\gt ; On the of Statisticat to the General of and Ether .
By J. H. JEANS , M.A. , Fellow of Trinity College , Cambridge .
( Communicated by Professor J. Larmor , Sec. R.S. Received May 19 , \mdash ; Read June 8 , 1905 .
) The General of al 1 .
One branch of abstract dynamics , which is , perhaps , best known under the name of Statistical Mechanics , attempts to discover as much as possible about the motion of a dynamical system when the specification of the system is either partially or fully known , but the initial configuration of the system is either not known at or is only partially known .
A complete solution of the problem would require not only a full knowledge of the dynamical specification of the system , but also a full knowledge of the initial configuration of the system .
These not being given , it is obvious that the problem cannot be fully solved .
The method of statistical mechanics is as follows : We divide up all possible configurations of the system into mutually exclusive classes , , A2 , , and calculate the corresponding classes of solution after time corre sponding to the initial configurations of classes .
Let us call the linal classes solutions .
In selecting the original classes .
we arrange that the values of any co-ordinate in any class shall differ so slightly from one another that the final values of the same co-ordinate in the solution shall also only differ slightly from one another .
The calculation proceeds by an appeal to the calculus of probabilities .
Let be the unknown probabilities that the co-ordinates of the initial system belong to the classes , so that Let us suppose that of the final classes , a certain number , , possess some special feature .
Then the probability that the system shall possess this feature after time is while the probability that it shall not possess this feature is Thus the odds in favour of the occurrence of this feature will be .
( 1 ) fft Mechanics to Dynamics of Matter and Ether .
Let us suppose that a certain number , , of the co-ordinates of the system all occur in exactly the same way in the dynamical specification of the system , so that the energy funct , ion is symmetrical in these co-ordinates .
Then the value of the fraction ( 1 ) will be a function of and of certain constants , which enter in the specification of the system and in the probabilities The procedure of statistical mechanics is to search for peculiarities for which the fraction ( 1 ) approximates to the limit infinity , when is very great .
Having found any such feature , we may say that it is infinitely probable that the peculiarity in question will present itself in the system after time this infinite probability being independent of the unknown law of probability of the initial co-ordinates .
As regards practical applications to material systems , an infinite probability of this kind may be regarded as a certainty\mdash ; e.g. , if the dynamical system is the universe , it is found to be infinitely probable that the entropy after time will be greater than the initial entropy , a theorem which , for practical purposes , is stated with sufficient accuracy by saying that the entropy of the universe continually increases .
ation to a Holonomic Dynamical System .
2 .
In what follows it is assumed that the system under discussion is " " holonomic\ldquo ; \mdash ; i.e. , th , its configul.ation at any instant is expressible in terms of a number of co-ordinates and their first diffelential coefficients with respect to the time , all these quantities being capable of independent variation .
It is almost unthinkable that in dealing with the ultimate structure of the universe , as it is proposed to do in the present paper , we should be concerned with a non-holonomic system .
Except in text-books , such systems can occur only through the massing together of a large number of independent co-ordinates into a smaller number , which nearly , but not quite , represent the larger number .
In fact , certain nces of independent co-ordinates are assumed to remain equal to zero absolutely throughout the motion , on the ground that , to make their difference vary perceptibly from zero , very great forces would be required .
The actual independence of the obliterated co-ordinates shows itself , physically as a ' force of friction and mathematically as an equation of constraint of the kind which is not permissible in a holonomic system .
We assume , further , that the changes in the co-ordinates of the system are governed by equations of the canonical form , ( 2 ) 3 .
If these pairs of co-ordinates are in number , we construct an imaginary space of dimensions , any orthogonal axes in this space corresponding to the values of the independent co.ordinates Any configuration of the system can be represented by a point in this space , and if the point is to continue to represent the system , it must move in the space with the determined by equations ( 2 ) .
If we suppose the space filled with a continuous fluid , we can represent simultaneously the differen motions of the system which correspond to all possible initial values of the co-ordinates .
If , in the notation already used , we arrange the fluid so that initially masses are found inside those parts of the space which represent the classes , A2 , .
and if this fluid moves as directed by equations ( 2 ) , then we shall find that after time't the regions of the space which represent classes occupied by masses of fluid .
The value of fraction ( 1 ) , the probability that the system shall now possess the special feature considered , is ( 3 ) 4 .
The first and most fundamental theorem of statistical mechanics expresses that the fluid moves as though incompressible .
For , by the equation of continuity , , and this vanishes by equations ( 2 ) , proving the theorem .
Thus there is no tendency for the fluid to accumulate in any parts of the spacc , and the volumes which represent the classes are equal to those of the corresponding classes , A2 5 .
If the full specification the system were known\mdash ; i.e. , if the form of the function , expressed as a function of the 's and , were known\mdash ; it might be possible , by eliminating the time from equations ( 2 ) , to obtain a number of integrals of the form constant , constant , etc. ( 4 ) Equations ( 4 ) not only represent integrals of the equations of motion : they also represent families of surfaces in the -dimensional space .
These families of surfaces have the property that the particles of fluid in the space 1905 .
] to Dynamics of Ether .
299 can never cross them ; the surfaces act as water-tight lpartmentS between which the fluid moves .
6 .
If there are no integrals of the form of equations ( 4 ) , function of in general vary with the time , and there will be no limits to the possible extent of this variation .
Let be any function , such that those parts of the space in which is numerically less than a certain finite value are finite in extent .
There are plenty of such functions to be found\mdash ; e.g. , energy , etc. Then , if we select a point at random from the whole space , it is infinitely probable that the value of will be reater than Suppose that we consider a system , or a point in the generalised space , starting from any initial value of , say , and moving for an infinite Knowing nothing about the initial co-ordinates of the System , we may say that all points in the space are possible points for the representation of ultimate state of the system .
Moreover , since by the theorem of S4 there is no tendency for the fluid to accumulate in aIJy regions of the space , we may say further that all points in the space have equal chances of representin .
the final state of the system .
Hence for a system starting from an arbitrary initial value of , it is infinitely probable that , after a sufficiently time , the value of will be Qoreater than any arbitrary value .
In other words , the tendency of any function possessing the properties we have ascribed to is to increase indefinitely in value .
In the terminology of S 3 , the special feature is that shall be positive ; the fraction ( :3 ) becomes infinite throu the number of terms in the denominator in comparison with the nutnber in the numerator .
Irrevcrsibility .
7 .
A tendency to increase such as that just is enerally described as an " " ible phenomenon We see at once that these irreyersible phenomena are inseparable from statistical mechanics .
The reversibility follows as a direct consequence of the absence of any tendency to on the part of the fluid in the generalised space , and this in turn is a consequence of the perfect reversibility of the original equations of motion .
In nature one function is known\mdash ; namely , the entropy\mdash ; which shows the tendency to increase indefinitely .
Other of the possible -functions do not , so far as we know , show this tendency .
The inference is that the hypothesis which we have been working , namely , that there are no compartments in the space , is not justified .
There ust therefore be at least one integral of the form of equations ( 4 ) .
One integral , as we , follows directly from the form of equations ( 2 ) We shall therefore suppose the generalised space dividedinto tight compartments by the family of surfaces constant , while these compartments , again , may be subdivided to any extent by surfaces representing other integrals of the system .
Since is necessarily positive , it is clear that must be infinite at infinity in the generalised space , the loci constant being closed or tubular surfaces , which , in general , decrease in size as the value of decreases .
The moving representative point no longer has access to the whole space , so that the argument of S6 breaks down , and the function no longer shows a tendency to indefinite increase .
8 .
The phenomenon of irreversibility does not , however , disappear altogether .
No matter how many integrals of the equations of motion there may be , it may still be true that in a single water-tight compartment some functions , , have the same value , , at all except an infinitesimally small fraction of all points inside the compartment , so that after the system has been in motion for an infinite time , it is infinitely probable that these functions will have the values .
Thus , the tendency of the functions , is always to move towards , never away from , the values , and this tendency represents an irreversible phenomenon .
9 .
A good illustration is afforded by the Theory of Gases .
In the development of this theory , it is found convenient at first to picture the molecules of the gas as rigid spheres , which together form a conservaGive dynamical system .
The co-ordinates of position and momentum are the rectangular co-ordinates of the centres of the spheres and the components of momentum of the spheres respectively .
The potential energy vanishes except during instants of collision , and these form only an infinitesimally small fraction of the whole time .
The kinetic energy may accordingly be treated as constant , and this kinetic energy is equal , except for a constant multiplier , to the sum of .
the squares of the momenta .
Let us examine the values of the momenta at each point of the space , and let us tabulate for each point of the space a function , such that the number of momenta at the point , of which the values lie within the narrow limits and , form a fraction of the whole .
Then the , which measures the value of averaged for all the momenta at the point , is a 1905 .
] Mechanics to of Matter and Ether .
definite function of the co-ordinates of the point , and possesses the property that at all except an infinitesimal fraction of the points of any compartment , its value is nil .
Thus , it is infinitely probable that in the final state of the gas the value of this function is nit , and as it can be shown that at all points at which the value of the function does not vanish , this value is necessarily positive , it follows that there is an irreversible tendency to a decrease of the value of the function .
When the value of the function reaches its final zero value , the gas obeys Maxwell 's law of distribution of velocities .
In a universe which consists of matter only , without ether , so that the energy is the sum of the of innumerable molecules , and of the potential energy of the intermolecular forces at work between the different molecules , it is easily shown that a similar irreversibility must be found , and that in the ultimate state of the system the energy is distributed in such a way at the average kinetic energy of each degree of freedom of each molecule is the same .
tical Mechanics of the Ether .
10 .
Let us next examine the statistical mechanics of a universe in which ether exists alone without matter .
The energy in this universe consists of the energy of trains of waves in the ether .
If the ether has any boundaries , these must not be regarded as capable of absorbing or storing energy , but as perfect reflectors impervious to energy .
Trains of waves may differ as regards wave-length and direction , and after the wave-length and direction of a train of waves are known , the train will be specified dynamically by the values of one pair of co-ordinates .
If we are considering a piece of ether limited by boundaries , we must replace the trains of waves by the principal vibrations of the ether within the boundaries .
In either case , the motion can be fully determined when the initial values of the co-ordinates are given , so that strictly speaking there is no need for the application of the methods of statistical mechanics .
It will , however , be of interest to examine what information is obtainable by these methods .
There is no interaction between the different trains of waves or principal vibrations , so that the energy of any train or principal vibration will remain constant .
The equations which express the constancy of this energy are equations of compartments in the generalised space , and if is the total number of co-ordinates , these equations are in number .
There are , however , other integrals to be obtained .
The energy in either case is supposed to be of the form VOL. LXXVI.\mdash ; A. where .
Thus and we have the integrals cons .
N-l in number .
Thus the total 1iumber of integrals is , [ and the compartments are reduced to lines .
A certain number of functions of the co-ordinates can be found which do not remain constant throughout the whole length of a line , but which have the same values at all except an infinitesimal fraction of the whole space .
As an example we have at all except an infinitesimal fraction of the points .
Thus , the method of statistical mechanics shows that the value of the function on the left hand tends to approach the value zero , i.e. , the electric energy tends to become equal to the magnetic energy .
This theorem , obvious enough from many points of view , is a theorem exactly analogous to the theorem of equipartition of energy in the theory of gases .
It enables us to see very clearly the kind of irreversibility with which we are concerned in statistical mechanics .
Statistical Mechanics of Matter Ether .
11 .
We shall next pass to the consideration of a system in which both ether and matter are present .
At first we shall suppose the interaction between ether and matter to be very slight , although not actually zero .
The advantage of this supposition is that the energy can be divided into energy of matter and energy of ether ; just as , in the Theory of Gases , a similar supposition justifies us in dividing the energy into the sum of the energies of separate molecules .
The whole system can now be regarded as nearly equivalent to the aggregation of two systems , namely , the matter and the ether , and its co-ordinates will be very nearly the same as the aggregation of two sets of co-ordinates , namely , those of the ether , neglecting the presence of the matter , and those of the matter , neglecting its interaction with the ether .
If , are 1905 .
] Mechanics to Ether .
the functions of the co-ordinates of the material system which remain constant when the ether is left out of account , and , the similar functions of the co-ordinates of the ether , then the whole system of quantities will nearly , but not quite , remain constant .
No single one of these quantities can remain absolutely constant , for if it did , there would have to be either modes of vibration of the ether which could not be influenced by the presence of matter ; a rather inconceivable supposition , or else it would be necessary that the ralues of the quantities which remain constant throughout the motion of the matter when the ether interaction is neglected should be uninfluenced by waves in the ether .
These latter quantities , however , are known to be the energy and the six components of momentum of the material system , and the phenomenon of the pressure of radiation shows that all these quantities are altered in value by waves in the ether .
The first possibility , however improbable , that there are ether vibrations which cannot be influenced by matter , need not trouble us .
Even if there are such vibrations they are by hypothesis incapable of affecting the motion in any way , and they will be totally imperceptible to us , for our only perception of occurrences in the ether is derived through their influence on matter .
One quantity , we know , remains constant through the motion , namely , the total energy of ether and matter .
This may be regarded as a modification of the equation which expressed , when matter alone was present , that the total material energy remained constant .
The question arises as to whether there are other constants which may be regarded as modifications of the remaining constants of the separate systems .
Although the material momentum parallel to the axis of does not remain constant when the interaction with the ether is taken into account , yet it may be that there is some function of the ether co-ordinates which may be treated as ether momentum , so that the -momentum of the ether plus that of matter may remain constant .
All that can be said on this question is that if there are such functions , they are entirely unknown to us , and that it does not seem likely that the neglect of the constancy of any such quantities will invalidate the conclusions to be arrived at .
We shall , therefore , proceed upon the assumption that the only constant is the energy .
12 .
It will simplify the problem to begin by considering either a finite universe , or else a finite portion of an infinite universe , enclosed within a perfectly reflecting boundary .
Let the of degrees of freedom of the matter inside this boundary , neglecting the interaction with the ether , involved in treating it as infinite will be negligible .
Let the of degrees of the ether be , giving co-ordinates to the aggregate system .
If we suppose the ether to have an absolutely continuous structure , the number will be absolutely infinite .
If , on the other hand , we suppose the structure of the ether to be such that when magnified sufficiently it is found to be in some way coalse-grained\mdash ; e.g. , if the ether were granular as imagined by Professor Osborne Reynolds\mdash ; then the number will be actually finite , although capable of being treated as infinite without perceptible error .
In either case , must be supposed large in comparison with N. The energy of the co-ordinates of the ether is , except for small terms arising from the interaction , expressible as .
sum of squares .
The energy of the material co-ordinates may , again ecting small terms , be divided into kinetic and potential energy .
The kinetic energy is expressible as a sum of squares , namely , the sum of the three components of energy of each electron of which the matter is composed .
This statement requires modification if any perceptible fraction of the electrons move with velocities comparable with the velocity of light .
We shall find it convenient at present to assume that the number of electrons moving with these high velocities is negligible .
Also , if the electrons are sufficiently near to one another , the kinetic energy must be modified by the introduction of cross terms\mdash ; products of two velocity components of different electrons\mdash ; but the energy still remains quadratic in the components of velocity , so that it can be at once transformed to a sum of squares .
Thus , the total energy is expressible as the sum of squares , plus an unknown potential energy of electrons .
It now follows , as in the proof of the well-known theorem of equipartition of energy , that throughout that compartment of the generalised space in which the system must now be represented , the sum of any of these squares stands to sum of the remaining squares in a ratio which is equal to at all except an infinitesimal fraction of the compartment , subject only to the condition that and are large enough to be treated as infinite without appreciable error .
Since and satisfy these conditions , it follows that the system tends towards a state in which the energy of the ether is infinite in comparison with the kinetic energy of the matter .
In other words , there is a general tendency for the ether to gain energy at the expense of matter .
It is , however , obvious that our own universe is at present far removed 1905 .
] to Dynamics of Matter and Ether .
305 from its final state , so that the study of this final state is of less interest than the study of the stages through which the final state is being reached .
13 .
In discussing the transition to the final state , a principle which I have proved elsewhere*will be of the greatest service .
Suppose that a vibration of any dynamical system is influenced by an external agency .
Then the principle in question asserts that the ultimate effect of this influence is infinitesimal , except when the external agency changes to a considerable extent in a time comparable with the period of the vibration .
If the time of change in the external agency is times the period of the vibration , where is large , then the ultimate change in the energy of the vibration vanishes to the same order as , a quantity which soon becomes ible as increases .
Thus , if is some small interval of time , so small that the material system may be arded as perceptibly unaltered through a time , then the change produced in the energy of ether vibrations of which the period is less than will be very slight .
The energy of such vibrations may therefore be treated as though it were incapable of change , at any rate so long as our consideration of the system does not extend over a very long period .
Hence certain of the ether vibrations , namely , those of shortest period , may be left out of account altogether , as has already been done in the case of the vibrations of the type discussed in S11 .
The total number of of vibration of any enclosed or unenclosed piece of ether is , as has been said , either very great or infinite , but the number of vibrations of an enclosed piece of ether of which the frequencies are below an assigned value is finite .
Thus , we can now suppose replaced by some smaller number Nf ' , and the value of will be finite .
So as we limit our consideration of the system to a finite time , say a million years , we may regard the equations expressing the constancy of the energies of the remaining modes of vibration as water-tight compartments in the generalised space , so that the whole motion may be regarded as taking place in a generalised space of only dimensions .
The ratio of ethereal to material kinetic energy which obtains through all but an infinitesimal fraction of this space is now , a quantity which cannot be infinite and may be very small .
It may be that there will be found to be other quantities besides the energies of ether vibrations of high frequency , which remain constant throughout enormous lengths of time .
This possibility will be discussed later , but meanwhile a few considerations suggest themselves in connection with the results just obtained , and these it will be well to discuss now .
" " The Dynamical Theory of Gases chap .
9 .
I have discussed the actual value of this ratio in another paper : " " On the Partition of Energy Matter and Ether ' Phil. Mag July , 1905 .
Mr. J. H. Jeans .
Application of Statistical Radiation at a Temperature .
14 .
In the first place , it must be remarked that the conception of the " " radiation corresponding to a given temperature\ldquo ; is not an exact conception .
It depends on two quantities besides the temperature of the bodies inside the enclosure , , the shape and size of the enclosure and the time which has elapsed since the system was enclosed .
For instance , it has been seen that after a system has been enclosed for an infinite time the radiation corresponding to any temperature is infinite , a result which is only reconciled with the finiteness of the total energy inside the enclosure by the temperature vanishing .
To simplify as far as possible , the question of the dependence of the radiation on the time , let us suppose that at the instant at which the system is enclosed , the whole energy resides in the matter .
If is a small time satisfying the conditions specified in S 13 , then the rate at which an ether vibration of high frequency gains energy will involve a factor , so that the time required for the vibration to acquire a perceptible amount of energy will involve a factor .
This is , of course , only true when is large .
At first , the energy of those vibrations for which is not large increases , and a state will soon be reached in which these vibrations have the share of energy allotted to them by the theorem of equipartition of energy .
With the progress of time the energy of the remaining vibrations gradually becomes perceptible , until ultimately the final state is reached .
15 .
We cannot , however , realise in nature the boundary impervious to all forms of energy , so that it is important to consider whether these predictions have to be modified if the boundary , instead of being perfect , is simply as perfect as we can make it .
The imperfections of a natural boundary be represented by supposing that the energy of the various ether vibrationsleaks out through , or is absorbed by , the boundary at a certain rate .
Let us suppose that the energy of ether vibrations of frequency inside the boundary , is per unit volume , and that the similar energy outside is per unit volume , then the best assumption to make is that the rate of leak is , where is a constant .
The rate of increase of is now given by the equation where is the rate at which the energy would increase if the boundary were perfect .
The solution of the equation is so that the ultimate value of is .
Experience shows that the value of must be supposed very small , while the value of the term 1905 .
] Mechanics to Dynamics of Matter Ether .
has been supposed to be small on account of the exponential .
Thus the energy of ether vibrations of high frequency tends towards a limit which is small in comparison with that predicted by the theorem of equipartition .
The value of the total energy per unit volume of all the vibrations of high frequency , tends to the value .
Now , , the energy per unit volume of these vibrations outside the boundary , is necessarily finite and is in point of fact known to be small .
The number of modes of vibration of the ether inside any given enclosure , of which the frequencies are large , and lie between two near values of and , is easily seen to be proportional to .
Since the integral is convergent , it follows that the total energy of these high-frequency vibrations tends to a finite limit .
Thus there is no longer any tendency for the energy of the matter to vanish in comparison with that of the ether inside the enclosure ; the two tend to assume a finite ratio , although neither of the actual energies can be permanent , as the system inside the enclosure is no longer a conservative system .
This definite ratio between matter and ether , however , lends a meaning to the expression " " radiation at a given temperature at any rate so long as we are concerned with the same enclosure and the same enclosed matter .
16 .
We may next discuss how this radiation at a given temperature will depend on the size of the enclosure .
So long as the temperature and the nature of the matter remain the same , we may suppose that all vibrations of frequency below a certain value receive their full share of the energy , while those of frequencies above receive a fraction of their full share , the amount of this fraction depending only on the value of .
Now , as the linear dimensions of the enclosure vary , the number of vibrations of which the frequencies lie within any given range of values will vary as the cube of the linear dimensions , and therefore the volume of the enclosure , provided the number in question is large .
* Thus , the total energy of the ether will be proportional to the volume .
If , then , we measure the intensity of radiation at a yiven temperature by the amount of ether energy per unit volume , we see that the intensity of radiation is independent of the size of the containing boundary .
It is also easy to see that it will be independent of the shape of the vessel , although less easy to see how to construct a satisfactory formal proof of this statement .
In a sense , then , it is permissible to speak of the ' temperature\ldquo ; of the ether inside the enclosure .
If the system of matter consists of a number I have verified this for a rectangular boundary ( Phil. Mag .
ante ) , and similar proof is easily seen to hold for a spherical or cylindrical boundary .
See also a letler by Lord Rayleigh , in ' Nature , ' May 18 , 1906 .
natural to think of the process as one analogous to conduction of heat , the ether in the final state possessing the same temperature as the bodies .
We now see that this conception receives a large amount of justification from statistical mechanics .
After agreeing , as a first approximation , to treat as invariable the energies of the rapid ether vibrations ( and , if necessary , th rapid vibrations of the material systems ) , we find that the representative point in the generalised space appropriate to the representation of the problem , is constrained to move in a compartment in which the only variable co-ordinates are those corresponding to the slower modes of vibration of the ether , and certain of the co-ordinates of the material systems , these latter always including those which represent the motions of translation of the matter , i.e. , those which measure the temperatures of the different bodies in the enclosure .
Throughout all except an infinitesimal fraction of the compartment , the mean energy of a material degree of freedom is equal to the mean energy of an ether degree of freedom , so that there is a tendency not only to the equalisation of the temperatures of the material bodies , but also of matter and ether .
We say then that the ether has a temperature , when this ether and a system of material bodies together in an enclosure have reached their final state , and this state is one in which the temperature of the matter is found to be T. On the other hand , when we are told that hydrogen or oxygen is at a temperature , we know the exact value of its kinetic energy per unit mass .
This is not so with the ether ; we know the energy of each of the graver vibrations , but we do not know how many vibrations have to be included in the total energy of the ether .
For instance , in a mass of hydrogen at temperature , the molecular movements are much quicker than those in a mass of mercury vapour at the same temperature , so that when an enclosure contains hydrogen , the number of vibrations which have to be considered in estimating the total energy of the ether is much greater than when the enclosure contains only mercury vapour .
Thus the intensity of radiation depends not only on the temperature , but also on the nature of the matter with which it is in equilibrium .
Stefan 's Law .
17 .
The next question which suggests itself is one of great importance .
Given the nature of the matter , in what way will the radiation vary with the temperature ?
Stefan 's empirical law states that the radiation is proporrional to the fourth power of the absolute temperature , and Bartoli and Boltzmann 1905 .
] Mechanics to Dynamics of and Ether .
309 have attempted to raise the law to the level of a theoretical law , which can be predicted by thermodynamical considerations .
The following is Larmor 's form of statement of the argument.* We have to suppose an interior body at temperature , surrounded by an exterior body A2 at temperature , but separated from it by a perfectly leflecting shell in the space between , which will prevent equalisation of temperature through passage of radiation from one body to the other .
The spaces on the two sides of the shell will each be filled with radiation of the constitution and density corresponding to the temperature of the body on that side .
We can imagine an ideal pump , constructed of perfectl .
material , that will pump radiation from the one side of this shell to the other , working against the difference of radiant pressure between the two sides : when the piston of such a pump is drawn out , the energy of the radiation that is isolated in the cylinder must be diminished by the work done by its pressure on the retreating piston .
The result will be that if and are the pressures of radiation on the two sides , then for each unit volume of radiation transferred by the pump from outside to inside , the outer body must emit energy of amount , made up of the ener , of the adiation and the work done by it on the piston , while the inner must absorb exactly what remains of this after the mechanical work is performed .
Now by Carnot 's principle , we have for such an engine working reversibly between temperatules T2 and In the present case , if the temperatures and on the two sides of the partition differ by a finite amount , the determination of the work will involve an integration ; let us , therefore , take the difference of temperatures to be infinitesimal , say , when the work will be equal to , or , to the first order .
As is or , we have thus which yields on integration const .
Thus we arrive at the empirical law enunciated by Stefan , that the density of radiant energy corresponding to any given absolute temperature is proportional to the fourth power of that temperature 18 .
The argument is seen to rest fundamentally upon the application of Carnot 's principle to the working of the heat here described .
The justification of Carnot 's principle may be legarded as either empirical or " " Ether and Matter p. 137 .
is material ; there is not , and cannot be , any evidence as to the working of an engine in which ether is used as working substance .
If it is argued that the principle is not known to fail in the case of anymaterial substance , and that there is therefore a strong argument by analogy that the ether not form a solitary exception , it may be replied that statistical mechanics explains quite definitely why the principle is applicable to all matel'ial substances , and proves with equal definiteness that it is not applicable to the ether .
We come to this question in discussing the theoretical justification , or want of ustification , for the principle when the working substance is ether .
Carnot 's principle is , in effect , identical with the second law of thermodynamics , and this in turn is a special case of a special proposition in statistical mechanics .
In the present investigation we have used the most general methods of statistical mechanics , and have arrived at a conclusion different from that of the second law .
The general investigation ought , of course , to take precedence over the attempted extension of the special case .
It is , moreover , easy to find the exact point at which the general argument parts company with that used in the special case .
In the special case , we are dealing only with forms of material energy such that there is an easy and rapid transfer of energy to the final state .
increase of entropy , as has already been remarked , indicates simply the tendency to move towards this final state , and Carnot 's principle is seen to be a special case of this general tendency in which it is supposed that the working substance is at every instant in the final state appropriate to its energy at that instant .
When the ether is in question , we have seen that the transfer of energy , instead of being infinitely.rapid , is , in point of fact , extremely slow , so that we never have to deal with a final state at all .
The way in which the argument suitable to the special case of a matel'ial working substance breaks down when applied to the ether will , perhaps , be understood clearly on considering an ether in which the special argument would be admissible .
Such an ether is an imaginary granular ether , in which the transfer of energy from matter to ether is extremely rapid , that the radiation at any temperature is simply propottional to that temperature .
19 .
At the same time , the theoretical justification usually given for Carnot 's principle is not one based on statistical mechanics , although it amounts ultimately to the same proof as that given by statistical mechanics , except that the theorem of equipartition is replaced by an appeal to experience in assuming Cf .
" " The Kinematics and Dynamics of a Granular Medium S 20 , 'Proc .
Lond. Math. Soc. ' ( 2 ) , vol. 3 , p. 124 .
1905 .
] to Dynamics of Matter Ether .
311 the tendency to equalisation of temperature .
We can easily find in this more usual proof , the particular assumption at which the two cases of matter and ether as working substances must part company .
It has to be assumed for the argument that the energy of the working substance is a function of only two independent variables , e.g. , the temperature and the density .
This is not true in the case of an engine in which ether is the working substance ; as we have seen , the ether energy is the sum of a number of vibrations of different wave-lengths , and the number of vibrations which have to be included in this sum will depend on the nature as well as on the temperature of the matter with which the ether is in communication .
Even if we avoid this objection , by considering only the cases in which all the matter is the same , we are left with a still more fundamental objection .
Our analysis of the energy of the ether into a sum of energies of separate vibrations has rested on the supposition that the boundaries of the ether rested fixed .
In the proposed argument for Stefan 's law , the piston of Ghe pump forms a moving mdary for the ether , such as has had to be specially prohibited .
The action of such a pump would change the frequency of vibrations in the ether , and energy at one instant to a vibration of one period would , after passing through the pump , belong to a vibration of some entirely different frequency .
The energy of the vibrations of high frequency no longer remains unaltered and very small , for there is a transfer of energy to these vibrations at every stroke of the pump .
Thus the existence of the pump may be regarded as .
down the compartments in the generalised space , and the only quantity which remains constant through a long interval of time is the total energy .
The system will rapidly assume the final state appropriate to the value of this total and this is a state in which the energy of matter vanishes in comparison with that of ether .
Thus Bartoli 's proof might be applicable to a universe in which pumps of the kind assumed had an actual existence , but has no application to our own universe in which the vibrations of highest frequency do not come into play at all .
Moreover , if Bartoli 's proof were applicable , the constant multiplying the factor would be infinite , so that we should have , except in a universe in which the energy per unit volume is It now be seen that in attempting to obtain a law of radiation in conformity with the analysis of the present paper , we shall not be able to use any method so general as that of the second law of thermodynamics .
In fact , all general methods of this kind apply only to systems which may be assumed to be in their final states , while to determine the actual radiation we have to discuss the path towards this final state .
The whole question is not one of partition of energy , but of transfer of energy .

