]\gt ; [ 239 ] VII .
Th Extinction of Sound in a Viscous Atmosphere by Smalt Obstacles of Cylindrical and Spherical Form .
C. J. T. SEWELL , B.A. , Trinit .
College , mbridge .
Communicated by Prof. H. LAMB , F.R. Received February l , \mdash ; Read March 10 , 1910 .
CONTENTS .
Page S1 .
ssion obtained for the dissipation function in the case of aerial vibrations . . . .
241 S2 .
Solution of the equations of motion with reference to cylindrical surfaces . . . . .
242 S3 .
Incidence of plane waves of sound npon obstructing cylinder .
An expression is ined for the secondary waves . . . . .
244 S4 .
Determination of the lost energy .
Discussion of resuIts . . . . .
248 S5 .
Incidence of plane waves of sound upon a large number of . . . .
254 S6 .
Problems reIating to spherical obstacIes .
A solution of the equations of motion suitable to such problems is obtained . . . . .
257 S7 .
Incidence of plane waves of sound upon obstructing sphere .
An expression is obtained for secondary waves . . . . . . . .
269 S8 .
Determination of the lost energy .
Discussion of results . . . . .
262 S9 .
Incidence of plane waves of sound tlpon a large number of spherical obstacles . . . . .
265 INTRODUCTION.\mdash ; The theory of the incidence of waves of sound in a non-viscous air upon small obstacles of cylindrical or spherical form is well known to students of mathematical physics ; it has been treated in Lord RAYLEIGH'S Theory of Sound , ' and in Prof ' Treatise on Hydrodynamics .
' The corresponding problems for a viscous air have not , however , been worked out , and this paper is devoted to an investigation of these problems .
The solutions of the equations of vibration of a viscous gas with reference to cylindrical and spherical surfaces were give by Prof in a paper titled " " On the Motion of a Viscous Fluid Contained a Spherical Vessel\ldquo ; and published in the Proceedings of the London Mathematical Society ' in 1884 .
It is easy to obtain solutions suitable to the case of divergent waves ; the functions involved are Bessel functions with a complex argument .
An analytical expression for the secondary waves diverging from the obstacle is obtained without difficulty .
It then remains to find an expression for the loss of energy to the ( 465 .
) 4.5.10 primary waves .
In calculating this loss of energy it is necessary to consider the dissipation of energy by friction in the immediate neighbourhood of the obstacle in addition to the energy which is carried away to a distance by the secondary waves .
This was pointed out to me by Prof LAMB , at whose suggestion this paper was written .
In obtaining an expression for the energy dissipated by friction I at first made use of the dissipation function .
This method led to exactly the same results as that finally adopted , but the mathematics involved were cumbrous , and the physical ideas , on which they were based , were somewhat obscure .
Another disad a of this method was that it was necessary to calculate separately the scattered and the dissipated energy .
I have to thank Prof LAMB for his kindness in pointing out to me the method of calculating the lost energy adopted in this paper .
The result has been to make the paper more clear and readable .
I have succeeded in obtaining expressions for the energy lost to the primary waves in the case of.spherical and cylindrical obstacles .
As might be expected , the problem of the cylindrical obstacle presents greater analytical difficulty than that of the spherical obstacle , and in the former case it is necessary to obtain different approximate expressions according to the diameter of the obstacle .
The results for wires of cm .
radius and for wires of cm .
radius can be obtained without much difficulty , but when the radius of the wire is of order cm .
it is necessary to perform very laborious calculations in order to arrive at intelligible results .
The energy lost to the primary waves is , in all cases , very great compared with what would be lost in a non-viscous air , but the ratio of the lost energy to that incident upon the obstacle is at most of order In the case of spherical obstacles the difficulties of approximation are not so great , a , s in the case of cylindrical obstacles the loss of energy is far greater than in a nonviscous air , but , as before , the ratio of the lost energy to that incident upon the obstacle is at most of order It is possible to extend the results obtained for a single obstacle to the case when the waves of sound are incident upon a large number of similar obstacles .
This has been done by Lord RAyLEIGH for the corresponding problem in a non-viscous air ; the same method has been adopted in this paper .
It should , however , be borne in mind that the results so obtained are valid only when the obstacles are so sparsely distributed that the space occupied by the obstacles.is a small fraction of the total volume .
The investigation has some practical interest .
It has been asserted that the suspension of a large number of parallel wires in a hall or lecture room will improve the acoustic properties of the room .
According to the theory of a -viscous air the effect of any such arrangement of wires would be inappreciable .
From the results of this paper it also appears that the viscosity of the air is not sufficient to accolmt for the alleged phenomenon .
The results in the case of spherical obstacles are of greater interest , since they are applicable to the consideration of the effect of foggy weather upon the propagation BY SMALL OBSTACLES OF CYLINDRICAL AND RICAL FORM .
and audibility of sound .
If the diameter of the drops of water in a dense fog is assumed to be 02 mm. , there is no appreciable alteration in the audibility of sound ; but , if the diameter of the drops of water is 002 mm. , the presence of fog is distinctly prejudicial to the audibility of sound .
The former case is in agreement with TYNDALL'S observations on the subject .
In conclusion I desire to thank Prof LAMB for much kind advice and encouragement in the writing of this paper .
S1 .
In a viscous gas , if , u ) be the components of the velocity at any point , of the fluid referred to fixed rectangular axes , and if be the pressure at this point , the equations of vibration may be written in the form ( 1 ) where is the equilibrium density , is a small constant of dinlensions depending on the viscosity , has been written for div , and terms of the order of the square of the velocity have been neglected .
By a method very similar to that*used in the case of an incompressible fluid it is found that the total rate of dissipation of energy within any closed surface is given by ( 2 ) where is the resultant velocity at any point of the fluid , denotes an element of the normal to the surface are the direction cosines of this normal drawn inwards in each case from the surface element SS .
Further , are given by the relations ( 3 ) .
When there is no motion of the gas parallel to the axis of , and the motion is the LAMB'S ' odynamics , ' p. 540 , VOL. CCX.\mdash ; A. 2 I same in all planes perpendicular to this direction , the expression ( 2 ) for the dissipation of energy per unit length of the axis of takes the form ( lv-mu ) ( 4 ) where is an element of the curve bounding the region in question .
S2 .
We now proceed to obtain a solution of the equations of motion which shall be applicable to the case when the motion in all planes perpendicular to the axis of is the same , and when further there is no motion parallel to this axis .
In this case the equations of motion take the form ( 1 ) , where , and denotes the operator The equation of continuity takes the form ( 2 ) , where squares of the velocity and other quantities of the same order have been ignored , and denotes the condensation .
If we neglect the effects of conduction and radiation of heat , we may write . . . . .
( 3 ) , where is the equilibrium pressure and is the velocity of sound .
Eliminating and from equations ( 1 ) with the help of ( 2 ) and ( 3 ) , we obtain ( 4 ) .
These equations will be satisfied by ( 5 ) , BY SMALL OBSTACLES OF CYLINDRICAL AND SPHERICAL FORM .
provided and are functions satisfying the equations ( 6 ) , .
( 7 ) .
If we assume a time factor , these equations take the forms . . . . .
( 8 ) , where and are given by , . . . . .
( 9 ) , ( 10 ) .
We shall for convenience suppose to be equal to that root of equation ( 9 ) which reduces to when is zero ; will be taken to be equal to With these conventions the solution of the equations ( 8 ) , which represents waves of sound div*erging from the origin , is giyen by ( 11 ) , where for convenience the time factor has been omitted and where is given by and have their usual significance as Bessel functions , and and are arbitrary constants .
It need hardly be remarked that to obtain the actual expressions for and it is necessary to multiply the expressions contained in ( 11 ) by , and to equate and to the real parts only of these products .
For the sake of brevity we shall usually omit , when possible , the time factor Since , in the case of air , is a small quantity , when expressed in cm .
sec. units , it is clear from ( 9 ) and ( 10 ) that for all audible sounds will be large compared with .
In fact , at any ordmary distance from the origin the terms in ( 11 ) will become insensible owing to the factor For plane waves propagated in the negative direction of the axis of , the solution will be given by . . . . . . . .
( 12 ) , where , as before , only the real part of is to be taken into account .
S3 .
of Plane of upon Obstructing Cylinder.\mdash ; We are now in a position to consider the effect of a cylindrical obstacle upon a train of waves propagated in a direction perpendicular to the axis of the obstacle and incident upon it .
We take the axis of the obstructing cylinder as axis of , and suppose the incident sound to be propagated in the direction along the axis of .
Then , as in ( 12 ) of the last article , we may assume for the incident sound the expressions Expanding in series of Bessel functions , we obtain ( ) ( 1 ) .
The scattered waves will be symmetrical about the axis of , or , and consequently we may assume for them the forms * .
( 2 ) .
At the surface of the obstacle the radial and tangential components of the velocity must vanish ; hence we must have ( 3 ) when , if is the radius of the cylinder .
In order that the boundary conditions ( 3 ) may be satisfied , we must have ( ha ) or ( ha ) ( 4 ) , and in general for ( 5 ) .
These equations ( 4 ) and ( 5 ) are sufficient to determine the various constants in the expressions ( 2 ) for the scattered sound .
In the process of approximating to the values of these constants by means of equations ( 4 ) and ( 5 ) , we shall confine ourselves to the case when ha is a small quantity ; in other words , we shall assume that the dimensions of the obstacle are small compared with the wave-length of the incident sound .
The other case , when the dimensions of the obstacle are large in comparison with the BY SMALL OBSTACLES OF CYLINDRICAL AND SPHERICAL FORM .
wave-length , presents exactly the same difficulties as occur in the similar problem in connection with the theory of a non-viscous gas .
We shall also consider especially the case when the viscous gas is the air of the atmosphere ; in this case is a small quantity about 132 in cm .
sec. units , and consequently we may regard as a small quantity for all wave-lengths .
Since is small , we may write very approximately from ( 9 ) , S2 , ( 6 ) .
We must now obtain approximations to the values of the constants in the expressions for the scattered sound by means of equations ( 4 ) and ( 5 ) .
For this we shall need the approximate values of the Bessel functions involved in these equations .
For convenience we shall write them down .
When is small , we have for all values of less important terms terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of terms containing higher powers of ( 7 ) .
Closer approximations are ( 8 ) .
Also we have , ( 9 ) .
Now , on eliminating from equations ( 5 ) , we obtain Further a well known result in the theory of Bessel functions .
Consequently we obtain ( 10 ) .
Now ha is a small quantity , and consequently we may use the approximations ( 7 ) ; substitutin .
for and from ( 7 ) , we obtain - or Hence we obtain as a first approximation .
( 11 ) .
Similarly , by elimination of between equations ( 5 ) , we obtain [ ha .
ka ( ha ) ( ha ) ( ka ) ] Using the approximate values given in ( 7 ) , we obtain I or ( 12 ) for all integral values of By a similar process , but carrying the calculation to a higher degree of approximation by means of ( 8 ) and ( 9 ) , we have .
( 13 ) .
Let us consider first the case when is small .
In this case it will be sufficient to derive the value of from ( 12 ) .
Hence we have approximately Writing for convenience , where . . . .
( 14 ) , BY SMALL OBSTACLES OF CYLINDRICAL AND SPHERICAL FORIsI .
and using the approximations for and given in ( 8 ) , we have Substituting these expressions for ( ka ) and in the above formula for we obtain In general , it will tend to sufficient accuracy if we take ( 15 ) in the case when is small .
Similarly we obtain from ( 11 ) the approximation ( 16 ) , when is smah1 .
Let us next consider the case when is large .
Since is great , we may write Substituting these expressions in the formula given for in ( 13 ) , we obtain approxinlately When is written for , this reduces to ( 17 ) .
As a first approximation to the value of , we find in case when is largs ( 18 ) .
The approximate value for , obtained from equation ( 4 ) with the help of ( 8 ) and ( 9 ) , is easily seen to be . . . . .
( 19 ) . .
Now that we have approximated to the values of the various constants involved in the expressions for the secondary waves , we can proceed to estimate the additional rate at which energy is being dissipated in the space surrounding the obstacle .
This additional rate of dissipation will be equivalent to the rate at which energy is being lost to the primary waves in consequence of the presence of the obstacle .
Now , if we consider a region bounded internally by the obstacle and externally by a cylindrical surface coaxial with the obstacle and of radius , it is clear that the difference of the rates at which energy is being carried across the internal and external boundaries of this region will be equivalent to the rate of dissipation of energy within it .
Hence , if denote the pressure and the radial velocity at any point due to the primary waves alone , and if denote the pressure and radial velocity due to the secondary waves alone , it is easily seen that the dissipation of energy within a distance of the obstacle is given by ( 1 ) , where it has been assumed that are both measured inwards , and the integration is to be taken round the boundary of the surface Now the dissipation of energy due to the primary waves alone is given by Hence it follows from ( 1 ) that the additional dissipation of energy due to the presence of the obstacle is given by ( 2 ) .
Now the rate at which energy is being carried across the surface is and hence from ( 2 ) it follows that the total additional dissipation of energy due to the presence of the obstacle is expressed by .
( 3 ) , where the integration is to be taken round the boundary of the surface Since is in all cases a small fraction , it is clear from ( 6 ) , S3 , that may be great compared with the wave-length of the incident sound and yet such that is a small fraction .
In this case we may neglect the imaginary part of in expressing the value of and at the surface .
Also , if is great , it follows that or is very great , since their ratio is the small quantity Since , when is great , approximates to the value * This method of finding an expression for the loss of energy was kindly suggested to me by Prof. LAMB .
I had previously obtained the same results by means of the dissipation function ; but the work involved was very cumbrous .
BY SMALL OBSTACLES OF CYLINDRICAL AND SPHERICAL FORM .
we see that the terms in the expression for the secondary waves are inappreciable at the external boundary owin to the exponential factor .
We shall suppose then that is great compared with the wave-length of the incident sound and yet such that is small .
Let us now return to the consideration of ( 3 ) .
At the external boun we may write approximately where square brackets are used to denote that only the real part of the expression so enclosed is to be considered .
Hence we obtain Using these expressions for and we obtain .
( 4 ) .
Again , since is great , we have from ( 1 ) , S3 , approximately Hence we obtain and Combining these expressions for and we find without difficulty Substituting in ( 4 ) and integrating with respect to we obtain VOL. CCX . .
2 of which the mean value is ( 5 ) .
This last expression , then , represents the loss of energy to the primary waves in consequence of the presence of the obstacle .
From the value of obtained in ( 12 ) of S3 we see that the summation ( 5 ) consists of a series of terms arranged in descending order of magnitude .
Consequently , in determining its value we shall limit our attention to the first two terms of the summation .
Hence the total loss of energy to the primary waves is given very approximately by . . . . .
( 6 ) .
Let us first consider the case when is small .
In this case we have , from ( 17 ) and ( 19 ) of S3 , Hence Now for small values of the radius the first term of this last expression is small compared with the second , and consequently may be neglected .
Hence the loss of energy to the primary waves is given approximately by ( 7 ) , in the case when is a small fraction .
The ratio of this last expression to , which represents the rate at which energy is incident in the primary waves upon the obstacle , is given by ( 8 ) .
We may notice from this last result that , when is small , the proportion of the incident energy , which is lost to the primary waves , is very nearly proportional to the reciprocal of the radius , since the logarithmic terms will change more slowly as the radius changes .
Hence , as in ( 7 ) , the total energy lost to the primary waves is almost ) endent of the dimensions of the obstacle , provided these are small enough to satisfy the conditions under which the results ( 7 ) and ( 8 ) have been obtained , and provided also that we limit our attention to obstacles whose dimensions are of about the same order of magnitude .
It might have been anticipated that the energy lost by friction in the neighbourhood of the obstacle would , in the case of very small obstacles , alter very slowly with the dimensions of the obstacle , and consequently BY SMALL OBSTACLES OF CYLINDRICAL AND SPHERICAL FORhI .
that the effect upon the primary waves of very small obstacles would be ahnost independent of the dimensions of the obstacles , provided they were ited to be of the same order of magnitude .
Let us now turn to the case when is great .
In this case we have approximately , from ( 17 ) and ( 19 ) of S 3 , Hence we find approximately Substituting this value for in ( 6 ) , we obtain , when is great , for the total loss of energy to the primary waves the formula The ratio of this to , which repl.esents the rate at which energy is incident upon the obstacle , is given by . . . . .
( 9 ) , which gives the proportion of the incident energy which is lost to the primary waves .
The first term in ( 9 ) is independent of the viscosity of the medium , and is obtained in the ordinary theory of a non-viscous air .
The second and third terms of ( 9 ) represent the additional loss of energy to the primary waves consequent the viscosity of the medium .
Further , since the ratio of the second to the third term of ( 9 ) is of order , it follows that the latter may be disregarded .
Hence we see that , since does not enter into the second term of ( 9 ) , the additional proportional loss of energy consequent upon the viscosity is almost independent of the magnitude of the obstacles .
In other words , the actual loss of energy in the pl.imary waves due to friction is proportional to the radius of ) obstructing cylinder if this be sufficiently large .
This last result is clearly what might have been expected on physical grounds .
It remains to consider the case when is neither very small nor very great .
In this case it is impossible to uate to the values of the Bessel functions involved .
From ( 13 ) S3 we have Hence it follows that where and square brackets are , as before , used to denote that the real part only of the expression so enclosed is being considered .
Hence we have Now , and consequently it may be neglected in comparison with Hence the loss of energy to the primary waves is given by The ratio of this to is ( 10 ) , which therefore represents the proportion of the incident energy , which is lost to the primary waves .
On p. 253 will be found a table giving the ratio of the lost energy to that incident upon the cylindrical obstacle in a number of different cases .
For wires and cylindrical rods of comparatively large radius it is necessary to use the formula ( 9 ) ; the results for wires of radii 10 cm .
, 1 cm .
, and cm .
have been deduced from this formula .
The formula ( 8 ) is applicable when the radius of the obstacle is very small , and the results for wires of radius cm .
have been obtained from it .
When the radius of the wire is of order cm .
, neither of these approximate formulae is applicable , and it becomes necessary to calculate the results directly from ( 10 ) ; the value of the ratio of the lost energy to that incident upon the wire has been worked out in this case for only a few values of the wave-length on account of the laborious nature of the work involved .
It should be added that in the table given on p. 253 denotes the wave-length of the incident sound measured in centimetres , and denotes the ratio of the lost energy to that incident upon the obstacle .
I have also worked out the values of for wires of diflerent diameters in the case when the wave-length of the incident sound is 250 cm .
The results are arranged below :\mdash ; BY SMALL OBSTACLES OF CYLINI)RICAL AND SPHERICAL FORM .
MR. C. J. T. SEWELL : EXTINCTION OF SOUND IN A VISCOUS ATMOSPHERE S5 .
Extension to the Problem of Number of \mdash ; Let us suppose that there are parallel wires per unit area of a plane perpendicular to their common direction .
At a distance from the axis of any one of these wires , great compared with the wave-length of the incident sound , the secondary waves due to that wire will be given very approximately in all cases by ( hr ) Since is great , we may write for all values of and takes the form Along the course of the primary waves this reduces to ( 1 ) .
Consider now the wires which occupy a thin stratum perpendicular to the course of the primary waves .
Let AP be the section of this Writing , we obtain In evaluating this last expression we may assume to be very small ; it then takes the form . . .
( 2 ) .
Now BY IALL OF CYLINDRICAL AND SPHFRICAL FORIsJ .
Also by a well known result we have Consequently Substituting this result in ( 2 ) and restoring the time factor , we obtain for the resultant at of all the secondary vibrations coming from the stratum . . . . .
( 3 ) .
When is small , is great compared with ; neglecting and using the expression for given in ( 15 ) , S3 , we obtain instead of ( 3 ) of which the real part is ( 4 ) .
To this is to be added the corresponding expression for primary wave .
The coefficient of is thus altered by the obstacles in the layer from unity to Thus , if be the energy in the incident waves , we have Integrating this , we obtain where is the energy in the primary waves at incidence and is given by ( 5 ) .
The coefficient of in ( 4 ) gives the refractivity of the medium as modified by the wires .
If S be the retardation due to the wires of the stratum Hence , if be the refractive index as modified by the wires ( 6 ) .
Hence we have ( 7 ) where denotes the ratio , assumed , of the volume pied by the wires to the total volume .
Let us now consider the case when is great .
In this case we have , from ( 17 ) and ( 18 ) S3 Substituting this expression for in ( 3 ) , we obtain for the resultant at of all the secondary vibrations coming from the stratum of which the real part is ( S ) .
To this is to be added the expression for the primary waves .
The coefficient of is thus altered by the obstacles in the layer from unity to Hence , if be the energy in the incident waves , we have Integrating this , we obtain where When and are substituted for and respsctively , this takes the form . . . . .
( 9 ) .
By SMALL OBSTACLES OF CYLINDRICAL AND SPHERICAL FORM .
The second term of ( 8 ) gives the refractivity of the medium as modified by the wires .
If S be the retardation due to the wires of the stratum Hence , if be the refractive index as modified by the wires , we have approximately . . . . . . .
( 10 ) where denotes the ratio , assumed small as before , of the volume occupied by the wires to the total volume .
If the waves of sound traverse a medium in which a number of parallel wires are arranged , then the reciprocal of will determine the distance which the waves will travel before the intensity of the sound is diminished in the ratio of .
For sound waves of wave-length 10 cm .
passing through a medium , in which there are 100 parallel wires of radius cm .
per unit area of a section perpendicular to the wires , this distance is 47 .
For greater wave-lengths the distance greater .
It seems hardly probable that any arrangement of wires could improve the acoustic properties of a room unless some other factor than viscosity is taken into account .
Of course , if is made sufficiently great , the reciprocal of may become very small ; but it seems probable that it would be difficult to arrange the wires so closely that , , , should be greater than .
If it was possible to arrange wires of radius cm .
so closely that ?
was , then the intensity of sound of waye-length 40 cm .
would be diminished in the ratio after passing through a thickness of less than 4 cm .
Such a contrivance could hardly , however , be carried out in practice .
S6 .
Problems to Spherical Obstacles.\mdash ; We require first a solution of the equations of motion suitable to such problems .
Differentiating the equations of vibration ( 1 ) , S1 , with regard to respectively and adding , we obtain with the help of ( 2 ) , S2 , . . . . . . . . .
( 1 ) .
If we now assume a time factor , this equation takes the form , where . . .
( 2 ) , ( 3 ) .
Also the equations of motion ( 1 ) S1 , may with the help of ( 2 ) and ( 3 ) , S2 , be written in the form ( 4 ) , where and ( 5 ) , ( 6 ) .
These equations ( 4 ) are satisfied by grad VOL. CCX.\mdash ; A. 2 where is any solution of the equation . . . .
( 7 ) .
The complete solution of the equations of motion will be givsn by grad . . . . .
( 8 ) , where satisfy the equations together with ( 9 ) .
The solution of these equations suitable to the case of waves diverging from the Ol.igin is given by ( kr ) ( kr ) ( kr ) ( 10 ) , with corresponding expressions for and .
Here and are solid harmonics of ositive degree , and ( kr ) is a function of given by the relation ( 11 ) , where and satisfy the relations ( 12 ) .
The general formulae for , , and , are ( 13 ) , ( 14 ) , ( 15 ) .
The functions , , and , all satisfy recurrence formulae of the types , ( 16 ) , ( 17 ) ; these will be found useful hereafter in reductions .
BY SMALL OBSTACLES OF CYLINDRICAL AND SPHERICAL FORM .
Returning now to the consideration of the equations of vibration , we find from ( 8 ) and ( 10 ) that the general solution suitable to divergent waves is given by ( kr ) ( kr ) ( ) ( 18 ) , with corresponding expressions for and here represents the general solution of ( 7 ) suitable to divergent waves .
Hence ( 19 ) , where is a solid harmonic of positive degree If the motion is in planes through the axis of , and is symmetrical about that axis , the solution takes the form ( ) ( 20 ) , with corresponding expressions for and .
Further we have ( ) , ( 21 ) , ( 22 ) , where and are arbitrary constants .
We write as before , where ( 23 ) , ( 24 ) .
From ( 15 ) it is seen that ( kr ) contains the exponential factor or ) ; consequently since becomes very great within a short distance of the origin , it is clear that , at a moderate distance from the origin , those parts of the expression ( 20 ) , which depend on the functions ( kr ) , become inappreciable and may be neglected .
Hence , at a sufficient distance from the origin , we may write grad , where ( hr ) .
S7 .
The Incidence of Plane Waves of upon an Obstruct , ng Sphere.\mdash ; We may now consider the effect of a spherical obstacle upon a train of plane waves of sound .
Suppose the centre of the obstructing sphere to be at the origin , and the sound to be propagated in the negative direction along the axis of ; then , as before , we may assume for the incident waves grad , where ( 1 ) , ( 2 ) .
Expandin in terms of the functions , we obtain ( 3 ) .
The scattered waves will be symmetrical about the axis of , and so we may assume for them grad ( kr ) grad ( kr ) grad ( ( 4 ) , where , ( 5 ) , ( 6 ) .
Now , by means of the recurrence formulae , it may be proved without difficulty that ( hr ) ( hr ) ( 7 ) .
At the surface of the spherical obstacle , we must have Hence , when / , we have grad ( ( kr ) grad ( kr ) grad ( Introducing the expressions given aboye for , we obtain for all values of when grad { grad { ( kr ) grad ( ( kr ) grad ( Hence , by means of the identity ( 7 ) , we find that the following expression ( ) and two other similar expressions must vanish , when , for all values of These three conditions will be satisfied if . . . . .
( 8 ) , BY SMALL OBSTACLES OF CYLINDRICAL AND SPHERICAL FORM .
and if for all values of ( 9 ) .
These equations ( 8 ) and ( 9 ) are sufficient to determine the various constants in the expression ( 4 ) for the scattered sound .
In the process of approximating to the values of these constants we shall limit ourselves , as before , to the case when is a small fraction ; in other words , we shall suppose that the radius of the obstacle is small compared with the wave-length of the incident sound .
We shall also suppose that the gas in question is the air of the atmosphere .
This will make a small fraction for all wave-lengths within the limits of audibility .
With this assumption we may write , as before , ( 10 ) .
Eliminating between the equations ( 9 ) , we obtain ( 11 ) .
Now with the help of a well known result in the theory of Bessel functions it may be proved that Hence equation ( 11 ) takes the form ( 12 ) .
Retaining only the principal term in the coefficient of , we may write approximately , whence we have . . . . .
( 13 ) , to the same degree of approximation .
Next eliminating between the equations ( 9 ) , we obtain ( ) ( 14 ) .
Hence by use of the expressions ( 13 ) and ( 15 ) S and , we have approximately , when is small , for all values of ( 15 ) .
We shall find it necessary to obtain a closer approximation to the value of Writing in ( 14 ) we have .
( 16 ) .
Now from ( 15 ) S6 we have approximately , since ha is small , , .
Substituting these expressions in ( 16 ) and making use of ( 13 ) S6 , we obtain - Writing ' for we obtain finally - - . . .
( 17 ) .
By a similar process we obtain from ( 8 ) , . . . . .
( 18 ) .
S8 .
Having obtained approximate values for the various constants involved in the expressions for the secondary waves , we may now proceed to estimate the additional loss of energy consequent upon the presence of the obstacle .
The method adopted is exactly similar to that of which we made use in the case of cylindrical obstacles .
As above , it is easily seen that the total additional dissipation of energy due to the presence of the obstacle is given by ( 1 ) where the suffixed letters have the same meaning as in S4 , and the integration is to be taken over the surface of a sphere of radius concentric with the obstacle .
As before , we shall suppose that is great compared with the wave-length of the incident sound , and yet such that is a smaU fraction .
By this assumption we enabled to neglect the imaginary part of and also to regard the motion as sensibly irrotational at the boundary BY SMALL OBSTACLES OF CYLINDRICAL AND SPHERICAL FORM .
Now , at the external boundary we may write approximately where square brackets are used to denote that the real part only of the expression so enclosed is to be taken into the accouYJt .
Hence we obtain Combining these results find .
( 2 ) .
Again we have from ( 3 ) S7 , since is large , Hence we have Combining the two last results we find Substituting this result in ( 2 ) and integrating over the surface of the sphere / we obtain of which the mean value is ( 3 ) .
This last expression then represents the loss of energy to the primary ways in consequence of the presence of the obstacle .
From the value of obtained in ( 15 ) , S7 , we see that the summation ( 3 ) consists of a series of terms arranged in descending order of magnitude .
Consequently , in determining its value we may limit our attention to the first two terms .
Hence the total loss of energy to the primary waves is given very approximately by . .
, . .
( 4 ) .
264 MR. C. J. T. SEWELL : EXTINCTION OF SOUND IN A yIscouS ATMOSPHERE Now , from ( 17 ) and ( 18 ) , S7 , we have . . .
( 5 ) .
Let us first consider the case when is small .
Since or is always small , it follows from ( 5 ) that in this case we may write approximately .
When is great , it is necessary to include one other term of ( 5 ) , and we may in general write in this case . . . .
( 6 ) .
Comparing this last result with that obtained in the case when is small , we see that we may take it as a sufficient approximation in almost all cases .
For small values of the radius the first term in ( 6 ) will be negligible .
Substituting from ( 6 ) in ( 4 ) we obtain for the total loss of energy to the primary waves the expression .
Now the energy incident upon the obstacle in the primary waves is given by , and hence the ratio of the lost energy to that incident upon the obstacle is ( 7 ) .
The first two terms of this last expression represent the proportion of the incident energy lost by friction .
The last term of ( 7 ) gives the proportion lost by scattering to a distance , and is the same as is obtained in the theory of a frictionless air .
When is small , the most important term of ( 7 ) is the second .
Hence we see that in the case of small obstacles.the ratio of the lost energy to the incident energy varies inversely as the radius of the obstacle , and consequently tends to become very great as this radius is diminished .
On the other hand , the actual amount of energy lost varies directly as the radius of the obstacle , and diminishes with the radius .
It is to be noticed that in the case of sufficiently small obstacles the energy lost to the primary waves is independent of the wave-length of the incident sound .
When is great , the most important term of ( 7 ) is the first .
Hence we see that in this case the ratio of the lost energy to that incident upon the obstacle is very nearly independent of the radius of the obstacle , provided the order of magnitude of this ratio is altered by the viscosity .
Consequently for sufficiently large By hIALL OF ] ) AN1 ) FORbI .
obstacles the loss of to the primary waves is proportional face of tlJe obstructing sphere .
There is , it will be noticed , a distinct similarity ) , these results and those obtained above in the case of cylindrical obstacles .
The has been evaluated in a number of different , and the are arranged on p. 266 in tabular form .
denotes the expreHHion ( 7 ) or the atio of the lost energy to that incident ) the obstacle , and represents the wave-length ( measured in centimetres ) of the incident sound .
S9 .
of tloe to of ) of ) .\mdash ; Let us consider now the loss of energy to the primary waves when these incident upon a large number of spherical obstacles .
We shall that there are small spheres per .cm .
; the validity of our al.gument will the volume occupied by the obstacles being small compared with the total Consequently ' must be a small fraction .
At a distance .
from of any one of these spherical } ) articles , compared with the of the incident sound , the secondary waves due that particle will be sensibly irrotational , and will be given approximately in all casss by iuce h is great , may and takes ; which , along the course of the waves , lceb t . . . . .
( ) .
Consider now the spheres which occup a thin stratum perpendicular to the course of the primary waves .
Let be any point in this stratum , and let ) the point where the vibration is to be estimated at a great distance the stratull .
If AP , the element of volume is and quelltly t of spherical particles in it is ) .
Also , if OP , AO , then and Now by ( 1 ) the resultant at of all the secondary vibrations which issue stratum given by d. Remembering that the angle AOP is to be regarded as very small , we see that the * See fignre , p. ) BV SMALL OF CYLTNDRICAL AND RICAL FORM .
resultant at of all the secondary vibrations coming from the stratum is given by . . . . . . . .
( 2 ) , where the time factor has been restored .
Now , from the results obtained in S7 for and , we find Except for very minute obstacles , it will be sufficient to write .
Substituting this last expression in ( 2 ) , we obtain for the resultant of all the secondary vibrations coming from tlje stratum - of which the real part is - ( 3 ) .
To this is to be added the corr sponding expression for the primary wave .
The coefficient of is thus altered by the obstacles in the layer from lmity to Thus , if be the energy in the incident way , we have Integrating this , we obtain where is the energy in the primary waves at incidence , and by .
( 4 ) . .
C. J. T. SEWELL : EXTINCTION OF SOUND IN A VISCOUS ATMOSPHERE If the radius of each obstacle is measured in centimetres , then the reciprocal of determined by ( 4 ) , will give the distance travelled by the sound before its intensity is diminished in the ratio of .
If the radins of each small sphere is cm .
, and there are per c. , then wil ] be a sulall fi.action , and the formula ( 4 ) will be licable .
With these nnmerical values we tain , in the case of sound of wayelength 50 cm .
, cm .
; hence , after passing through a thickness of less than 12 cm .
of such a medium , the intensity of the sound will be in the ratio of The formla ( 4 ) should be ) to fogs , as we may regard the water particles fixed , since their inertia is so much greater than that of the surrounding .
I indebted to Prof for the following information from HANN'S ' Meteorologie ' : " " In a dense the amount of water may vary from about 3 to 10 .
per cubic metre .
Assuming that the diam of the drops is 02 mm. , and a cubic metre contains .
of water , this is calculated to give drops per cubic metre , and therefore 10 per cubic centimetre With these numerical data the form ula ( 4 ) gives metres , and consequently it follows that the fog would not interfere ppreciably with the propagation of sound .
But if the diameter of the drops could be as small as 002 mm. , a fog of the same density would contain drops per cubic , and would be nearly equal to metres , and consequently the sound would be damped very quickly by the fog .
* On the othel ' observations appear show that the presence of fog is not ejudicial to audibility of sound .
The coefficient of in ( 3 ) gives the refractivity of the medium as modified by the spherical particles .
If S be the retardation due to spheres of the .
Hence , if be refractive index of the medium as modified by the particles , , where ) denotes the ratio , } small , of the volume occupied by the particles to total volume .
1-Ience finally we have ( 5 ) .
For sound of cm .
incident upon a medium in which there are spherical particles per cubic centimetre , each of radius cm .
, we obtain See , however , note at end .
BAYLEI , 'Treatise on Sound , ' Vol. II .
, p. 137 .
BY SMALL OBSTACLES OF CYLINDRICAL AND FORbI .
Note , April 12th , ] .
910 .
Prof LARMOR has kindly pointed out to me that it is not itimate to ) pply the formula ( 4 ) to fogs without further tion .
Although of the water particles is so much greater than that of the air , yet in consequence 01 the viscosity of the air it does not follow that we may regard the water particles as approximately fixed .
I investigated the problem of a free ) stacle .
The analysis is very similar to that in the problem of the fixed obstacle .
The secondary waves from the obstacle affected only in the terms containing spherical harmonics of the first order .
If is the velocity of along the axis of we obtain . . . . .
( 1 ) , ( 2 ) , ether with . . . . .
( 3 ) where is the component of the stress across the surface of the obstacle in the direction of the axis of and the integration is taken surface of the obstacle .
The last equation reduces to . . . . .
( 4 ) where is the density of the obstacle .
Eliminating and between the equations ( 2 ) , and ( 4 ) we obtain mately when is small Hence , if be the ratio of the amplitude of the motion of the obstacle to that of the waves of sound , we have . . . .
' . . . .
( 5 ) .
Again , eliminating and from ( 1 ) , ( 2 ) , and ( 4 ) we obtain approximately ( ) Hence we obtain without difficulty when is small . . . . .
( 6 ) , where denote the real part of the value of A in the case ) the fixed and free obstacle .
EXTINCTION OF SOUND IN yISCOuS ATMOSPHERE , ETC .
The last relation may be written in the form . . . . .
( 7 ) .
It follows that the ratio of the lost energy to that incident upon the obstacle is given by . . .
( 8 ) .
Extending this result to the case of a number of free spherical obstacles we obtain , instead of ( 4 ) , S9 where represents the ratio of the lost to the incident energy given in ( 8 ) .
It follows this investigation that the results obtained in this paper are only applicable to fogs if is a small fraction .
This condition is satisfied for obstacles of radius cm .
, and also for obstacles of radius cm .
when the wavelength of the incident sound is not too long .
In the case of obstacles of radius cm .
, however , this condition is no longer satisfied ; approaches close to unity for all wave-lengths , and consequently and are very small .
Hence , if the diameter of the drops of water in a fog is as small as 002 mm. , such a fog does not interfere appreciably with the propagation of sound , and a result is obtained in agreement with TYNDALL'S observations .
I append a table giving the values of in a number of different cases .
When the wave-length of the sound is very great , or when the obstacle is extremely minute , the obstacle vibrates with the air surrounding it . .
L. .
L. cm .
cm . .
L. 80 284 ] 971

