I. On Poisson 's Solution of the Accurate Equations applicable to the Trans-mission of Sound through a Cylindrical Tube ; and on the General Solution of Partial Differential Equations .
" By R. MOON , M.A. , late Fellow of Queen 's College , Cambridge .
Commnunicated by Prof. J. J. S-YLVESTER .
Received INovember 14 , 1866 .
( Abstract .
) The pair of equations v=(Pf{ ( v+a)t which constitute Poisson 's solution of tih accuralte equations applying to the transmlission of sound through a cylindrical tube derived by La Grange 's method , have long attracted the attention of mathematicians .
For La Grange 's equations we may substitute the following , viz. cv a2 dp a dt JD dx dv+D dp=o dx p2 dt The first of these is obtained from the equation of the Encyc .
Met .
( Art .
" Sound " ) ; by putting v for die and D for die dt p dx The second results from a similar substitution in the analytical condition , d[d ?
q d/ die Poisson 's solution has always been regarded as imperfect , and may easily be shown to be so .
I was some time ago struck by finding that the above equations , while they yielded with facility the result of Poisson , notwithstanding their simplicity , baffled every effort to extract from them one more consonant to the general exigencies of the problem .
I had at this time arrived at the conclusion that the law of pressure assumed in the received theory of Fluid Motion could not be generally true , and in a Paper* communicated to the Royal Society , had pointed out that , in a certain case of motion , the assumption of the truth of that law led to a contradiction ; while in another case of motion the expression for the pressure given by the received theory was palpably erroneous .
It occurred to me , therefore , that the defective law of pressure of the received theory accounted for the defective solution which alone was obtainable from the equations of motion derived from it .
If the law of pressure of the received theory was not always true , if it held only when certain conditions were satisfied , those conditions would obviously have the effect of dismissing from the complete solution of the problem obtained on a perfect theory at least one of the two arbitrary functions which it must necessarily involve .
With a view to establishing this point , assume the solution of the equations of motion to be contained in the pair of equations , F(xytpv ) = { f( ytpv ) } , .
( B ) F(tytop)4'={/ ( Z3)}* Jyv ee If these equations satisfy the equations ( A ) , the latter will equally be satisfied by the pair of equations , F(xytpv ) =v , F(wy tpv ) ={f(xy tpv ) } .
But it is shown in my Paper that these latter can only satisfy the equations ( A ) on the supposition that F is of the form F==F(v alog , p ) .
Moreover , since we have in equations ( B ) the function F equal to an X On the Trtue Theory of Pressure as applied to Elastie Fluids in Motion .
2c2 arbitrary function of f , conversely we must have f equal to an arbitrary function of 1 .
I-ence , in exactly the same way in which it is proved that F must have the above form , we may equally prove that f , F , f must severally have the same form .
It clearly results , therefore , from the above assumed solution ( B ) , that the equations ( A ) are insoluble , except on the assumption that the velocity may be expressed in terms of the density alone ; i. e. that we may assume v =f( ) .
But substituting this value of v in equations ( A ) , the latter become dp a dp ( p ) Sz2 + D , / dt D 4 , '_ f ( p p+ D dt)_o dx P dt-whence we get ) ' .
2=p dpf_ ap2 1dp .
/ ( ^ p ' ~di\ dr\ and eventually , D being the value of p when v=O , +h v_ loc. -D P ?
=- . ?
a t } which is the most general solution of which the equations of motion are susceptible ; and which , making allowance for the difference of the ordinates employed in the two cases , y referring to the particle when in motion , and x to its position of rest , is identical with that of Poisson .
The failure of mathematicians to derive from the equations of motion of the received theory a solution containing two arbitrary functions has hitherto , I apprehend , been universally attributed to difficulties of integration .
So far is this view from being well founded , however , that in a postscript to my Paper it is shown that , assuming the pressure to follow any law whatever , a solution of the equations of motion can be obtained containing two arbitrary functions ; a result , however , which requires that the expression for the pressure shall satisfy certain conditions , which conditions are violated when the pressure is assumed to vary with the density alone .
Whatever be the law of pressure , it must always be capable of being expressed in terms of x and t. Moreover , the velocity and density are in like manner severally capable of being expressed in terms of x and t ; whence it follows that we may always express x in terms of p and v , and equally that we can express t in terms of p and v ; so that , whatever be the law of pressure , we may assume it to be a function of p and v. Hence , assuming dp_dp , +dp dv , _p+ dv dx dp dax dv dx dx dx ' where R and V are functions of p and v only , we I have for our equations of motion the following , viz. dv R dp V dv ?
Ct +D+b D ' &amp; +dv D dp dx T2 dt of which the following pair of equations constitute the solution , viz. f2(p , v)=~ { V2 V2 -4Rp2 } Ae , ( p , ?
) 2V ' ' wheref , ( p , v ) , =const .
, and f , ( p , v)=const .
are the respective integrals of the ordinary differential equations , dv-_ W +4R dpdv- ?
p 2 --dp = O , 2p which involve the variables v and p only .
But this result is depellent on the fact of the following conditions being satisfied , viz. , that we have 3dp dv where K , =VVV2+411p2 , K , =VV4Rp2 , which conditions cannot be satisfied if the law of pressure depends upon the density only ( in which case =0 and R contains p only ) , as may easily be shown .
With regard to the theory of Partial Differential Equations , I conceive that the methods indicated in the Paper will serve to elicit every solution of a partial differential equation of the second order and first degree , save one , viz. , an integral solution consisting of a simple relation between the variables x , y , z , free from arbitrary functions , and which is not derivable from a solution containing arbitrary fimctions , by assigning particular values to the latter .
If the equation O=Rd2 + dz +TV d ; ?
dxdy die have a first integral consisting of the pair of equations F(xyzpq)=of{f(xy-pq ) } , F(s yq ) =44 ( ytzIq ) , or consisting of the pair of equations F(xyzpq ) =0 , P( yzp= ) = { f(x yp , q ) } , or of the single equation F(yzp q ) = &lt; / f(xJyzpq ) } , or of the single equation F(xyzp ) =0 ) then in every one of these cases we must have 0==R .
F'(q)-S.(q ) .
F'().(p ) + T. '(p ) , O-R .
F'(x ) .
F'(q3 T. F'(29 ) .
F'(q ) + R. F'(q).2 + T. '(p ) ?
}F'(z ) +Y .
F'(p).F'(q ) .
It is also shown with more or less of generality , and it is capable of being shown generally , that if the given equation admit of a complete integral solution containing two arbitrary functions , it will necessarily have two first integrals , each of which will be of the form F(xypq ) = { f(xyz)q } .
It might have been added , that if the general equation of the second order and second degree be written O= Pr2 . ?
P , .
*2 2 ' t2 +P sP P. et + .
st +Pyr+P.S .
8+Pe +P ; and it is satisfied by the equation F(yZ-pq ) ( { f(xypq ) } then F and f must severally satisfy all three of the following equations , viz. F'(p ) ' ( ) =0 , F'(x ) qr':(O , z , , , Y'p ) = 0 , F'( ) + F ' , ( )~ -m '( ) =0 , where 0-P~ , .
14--_P .
13 ( Ps2 +Pg.)1--Ps .
lP , 0P nm2P , ,p 4-IP~2n2-P-P nP , 0-(2Pr--P2s .
l+P./ 12)M +(2Pe 12-Ps .
tPr)n -(P2_ Ps I+Pr ) .
T ' he functions f/ , m , must satisfy the sa-ne conditions , except in the second of the above cases .

