Normal Series satisfying Linear Differential Equations .
339 j " On the Normal Series satisfying Linear Differential Equations .
" By E. Cunningham .
Communicated by Dr. H. E. Baker , F.B.S. Received December 8 , \#151 ; Read December 15 , 1904 .
( Abstract ) .
If y denotes a column of n elements , and is a square matrix of | n rows and columns of elements , each of which is a function of the independent variable , n independent solutions of the system of 1 simultaneous equations die / dt \#151 ; uy are given by the n columns of the matrix 12 ( u ) = 1 + Qu + Q ( uQu ) + Q { \#171 ; Q ( wQ ?
t ) } + ... .
; where Q ( \lt ; f\gt ; ) denotes j \lt ; f\gt ; dt .
J ^0 In the 'Proceedings of the London Mathematical Society , ' vol. 35 , pp. 333 ff .
, the form of this solution is developed by Dr. H. F. Baker , for the particular case in which each element of u has a pole of order unity at the point t\#151 ; 0 .
This is the case of a system derived from a linear ordinary equation of order , all of whose integrals are regular in the neighbourhood of t =4 ) .
If x is any square matrix of order n , and a square matrix satisfying the system of equations dgjdt = ug gx\gt ; | then y \#151 ; g 12 ( x ) is a matrix satisfying die / dt \#151 ; uy .
This result is used by Dr. Baker in the paper above mentioned to show that in the above case a solution may be found in the form : ^12(X ) , where the elements of g are- simply series of powers of t , and the matrix 12(x ) is expressible in finite terms .
The form of the result puts in evidence what are known as Hamburger 's sub-groups of integrals .
The present paper deals with the linear system of equations in which the matrix u is of the form K +fp + * ' ' * + a ' + ^i are , fir being matrices of constants .
The general linear ordinary equation of order n which has 0 as I a pole of its coefficients is reduced to this form .
It is shown , in the first place , that provided the roots of the determinantal equation The Flow of Water through Pipes , \#171 ; The Flow of Water through Pipes.\#151 ; Experiments on Stream-line Motion and the Measurement of Critical Velocity .
" By H. T. Barnes , D.Sc .
, Assistant Professor of Physics , and E. G. Coker , M.A. , D.Sc .
, Associate Professor of Civil Engineering , McGill University ' .
Communicated by Professor Osborne Keynolds , F.R.S. Received November 4 , \#151 ; Read November 24 , 1904 .
In a brief note published in the ' Physical Review , '* we have described a thermal method of observing the change from stream-line to eddy motion for water flowing through pipes of different diameters .
This method we have recently employed in a series of experiments , which we have carried out on a much larger scale than was previously attempted .
The method had several points to recommend it for our particular work , the chief one being the simplicity of the experimental arrangements .
The object in undertaking this investigation , for which our method was adapted , was in the interest attached to a study of stream-line flow , where every form of initial disturbance was as completely removed from the water as possible .
The impossibility of heating a column of water uniformly throughout while flowing in stream-line motion has been previously observed . !
It was shown that , when water is heated electrically while flowing through a tube of two or three millimetres in diameter by a central wire conductor , the heat is carried off by the rapidly moving stream , which forms a cloak of hot water around the wire , and leaves the walls of the tube almost entirely un heated .
If the heat be applied to the outside of the tube , the heated water remains in contact with the inner surface of the tube and the water flows through the centre of the tube at a much lower temperature .
It is quite different if the flow is increased sufficiently to break up the stream-lines .
In this case , eddy or sinuous motion results , and the distribution of heat throughout the water column becomes uniform .
The change from stream-line to eddy motion can be very clearly observed in a tube heated on the outside , since the temperature of the emerging stream immediately increases when the flow rises above the critical point .
The point of change is very sharp , and the disappearance of the stream-lines instantaneous .
At the outset of our work we were guided by the two very important papers of Osborne Reynolds , !
in which the laws governing the flow of water in pipes and channels are worked out .
It is clear * ' Phyeical Review , ' vol. 12 , p. 372 ( 1901 ) .
f H. T. Barnes , ' PKil .
Trans. , ' A , vol. 199 , p. 234 ( 1902 ) .
X " Phil. Trans. , 1882 and 1895 .
342 Profs .
H. T. Barnes and E. G. Coker .
[ Nov. 4 from a study of the work of Reynolds that the change from streamline to eddy motion may take place within a wide range of velocities .
We may observe the critical velocity in two ways : either by observing the velocity at which the stream-lines break up into eddies , or by obtaining the velocity at which the eddies from initially disturbed water do not become smoothed out into stream-lines in a long uniform pipe .
The first change may be at any velocity within certain limits depending on the initial steadiness of the inflowing water , while in the second , the change can take place at only one velocity .
It therefore depends on whether we start with initially quiet water or disturbed water what value will be attained .
Below what we will call the lower limit of critical velocity or stream-line flow , the stream-lines will form the stable flow whatever may be the condition of the water before entering the pipe .
If it be in a disturbed state , a short length of pipe is required before the eddies are smoothed out , but streamlines finally appear further on and subsequently persist as the stable flow .
The greater the initial disturbance in the water , the greater will be the length of pipe probably required before the eddies disappear .
The flow at which the eddies persist throughout the entire pipe , however long , indicates that the change has taken place from ono kind of flow to the other .
This point of change is the true critical velocity , and although the production of eddies at the mouth of the pipe may vary , it will be independent of them .
If the initial disturbance , however , is not sufficient to prevent the water from starting in stream-line motion , the critical velocity is raised .
The change in flow in this case takes place by the birth of eddies in the pipe itself , and the point of change is in some way related to or limited by the degree of steadiness in the water .
In regard to this point , Reynolds says , on p. 955 of his original memoir : " The fact that the steady motion breaks down suddenly shows that the fluid is in a state of instability for disturbances of the magnitude which cause it to break down .
But the fact that in some conditions it will break down for a large disturbance , while it is stable for a small disturbance , shows that there is a certain residual stability so long as the disturbances do not exceed a given amount .
" The cmly idea that I had formed before commencing the experiments was that at some critical velocity the motion must become unstable , so that any disturbance from perfectly steady motion would result in eddies .
" I had not been able to form any idea as to any particular form of disturbance being necessary .
But experience having shown the impossibility of obtaining absolutely steady motion , I had not doubted but that appearance of eddies would be almost simultaneous with the condition of instability .
I had not , therefore , considered the disturbances except to try and diminish them as much as possible .
I The Flow of Water through Pipes .
1904 .
] I had expected to see the eddies make their appearance as the velocity increased , at first in a slow or feeble manner , indicating that the water was but slightly unstable .
And it was a matter of surprise 1 to me to see the sudden force with which the eddies sprang into \gt ; existence , showing a highly unstable condition to have existed at the time the steady motion broke down .
" In this connection it is a " matter of interest to note that when all forms of initial disturbance , as well as the disturbing influence of the walls of the pipe , are removed , eddy motion is no longer possible .
Such a condition of affairs we have in a jet of water issuing from a circular orifice .
It is easy to show experimentally that the beautiful rod-like appearance of the water jet depends on the absolute quietness of the water feeding the jet , and that the jet is immediately broken up by producing eddies artificially .
Jets flowing in stream-line motion at a speed of from 20\#151 ; 30 feet per second have been obtained by us from a two-inch orifice .
In our present series of experiments we arranged that our pipes should feed from the same tank from which we produced the jets , in order that we might be sure to have the water in a perfectly quiet state .
We could , therefore , be fairly safe in assuming that the birth of eddies was the result of the disturbing influence of the walls of the tube only .
Two methods of study were adopted by Reynolds in his experiments .
The first was the method of colour bands , which gave the point at which stream-lines break up into eddies , and which we have called the upper limit of critical velocity , and the second was a method of pressures , where the relation between velocity and pressure was obtained above and below the critical velocity .
In the colour-band tests a long narrow tank was used , in which the pipe was placed .
A flare , carefully smoothed , was fastened to the inflow end , while the outflow end protruded out of the tank and connected with the waste pipe .
A small tube containing coloured liquid was directed just in front of the flare and a thin stream of colour was drawn into the tube with the water .
This thread of colour remained intact all along the tube for stream-line motion , but disappeared with great suddenness when the critical velocity was reached .
The water in the tank was allowed to become as quiet as possible before beginning the experiment .
The rate of flow was calculated from the rate of lowering of the water surface , and the dimensions of the pipe .
In the pressure experiments the water was allowed to flow directly from the mains into the pipe .
After a sufficient length of pipe to allow of the dying out of the eddies , pressure gauges were attached at a fixed distance apart , and readings | of velocity and pressure made .
The critical velocity was not observed I directly by this method , but since the resistance changes with the I change in the flow , from the first power of the velocity to about the 344 Profs .
H. T. Barnes and E. G- .
Coker .
[ Nov. 4 l'7th power , the intersection of the logarithmic homologues gave a method of determining it .
Reynolds remarks that for a short distance above and below the critical velocity the gauges became unsteady and no readings could be made .
By means of his experiments Reynolds was able to verify his mathematical deductions , and he showed from three different pipes , together with a comparison with the experiments of D'Arcy for large pipes and with Poiseuille 's for small tubes , that the critical velocity varied inversely as the diameter of the pipe .
He further showed that the critical velocity followed the viscosity temperature law as deduced by Poiseuille , and , therefore , varied as the viscosity over the density .
Similar experiments with three different pipes by the method of colour bands showed that the upper limit followed the same laws approximately .
The theoretical equations , however , had to do with the lower limit .
In our experiments on the upper limit of critical velocity with absolutely quiet water , we found that stream-lines were preserved in many cases to very much higher velocities than would be expected from the inverse-diameter law , and that the upper limit falls off ' slightly more rapidly with rise in temperature than the law of Poiseuille .
We do not wish to lay stress on these points , or intimate in any way that we think they question the validity of the theoretical laws which have been so completely worked out by Reynolds .
We think , however , that they show the instability of the upper limit of stream-line flow , and how much it depends on forms of disturbance .
We have , moreover , made independent determinations of the temperature variation of the lower limit , and find by two different methods that over a wide range in temperature the critical velocity follows the viscosity temperature law accurately.* The slight deviation from the theoretical law for the variation with temperature of the upper limit which we have observed in the case of two pipes of different diameters is , we think , due to the fact that it becomes more and more difficult , as the temperature rises , to maintain stream-line flow in the unstable region .
This would result in the upper limit apparently falling off more rapidly than the temperature formula would allow for .
J The question has suggested itself to us that some form of initial disturbance in our tank may have come in as the temperature was raised to cause a greater falling-off in the upper limit than the theoretical .
Convection currents might have supplied such a form of disturbance , but the size of our tank and the fact that the mouth of our pipes was always* placed at or near the middle , together wit * E. Gh Coker and S. B. Clement , ' Phil. Trans. , ' A , rol .
201 , p. 45 ( 1903 ) ; H. T. Barnes , \#171 ; B. A. Report , ' Belfast ( 1902 ) .
\#163 ; 904 .
] The Flow of Water through Pipes .
345 the length of time required for the tank to cool down , make it appear to us improbable that convection currents played an important part .
Thermal Method of Measuring Critical Velocity .
In some of our first experiments , we observed the change in temperature in the column of water at the critical velocity by noting the increase in resistance of a platinum wire threaded through the centre of the tube , which was heated on the outside , and our preliminary results showed that the presence of a wire of 6 mils , thickness in a tube of \#163 ; inch in diameter was quite sufficient for our purpose .
We found that the point of change in flow could be observed more simply by placing the bulb of a mercury thermometer in the stream of water as it flowed out of the tube .
A glass prolongation , of slightly greater diameter and connected carefully to the brass pipe by a specially constructed cone or adapter , enabled the readings on the thermometer to be observed .
It was a matter of interest to us to see the sudden way in which the reading on the thermometer indicated the point of change by an almost instantaneous change of reading .
We were fortunate in having at our disposal , through the kindness of Dr. H. T. Bovey , F.R.S. , the facilities offered by the hydraulic laboratory , where the large experimental tank , 20 feet high and 25 square feet in area , served admirably for a reservoir .
The tank stood on the bed-rock , and was therefore free from disturbance in the rest of the building , and , after the eddies occasioned by filling had died out , the water was in a very quiet state .
The water used for the experiments was supplied from the Montreal mains and was quite clear ; and every precaution was taken , by repeated cleaning , to have the water and tank clean .
The action of finely divided matter in suspension in the water is to cause a breaking-up of the stream-lines , so that it was necessary for us to avoid this form of instability .
Our preliminary experiments , made to test the method before applying it on a large scale , were quite satisfactory , and , after profiting by various trials , the final apparatus took the form of that represented in fig. 1 .
In one side of the large tank A , a hole was cut , in which was fitted a special form of stuffing box to take the tubes of different sizes .
Each tube was made of brass and joined to a metal trumpet flare , such that the inside was filed perfectly smooth , and free from sharp points which might disturb the flow .
The flare was located at about the centre of the tank , so as to be as free as possible from convection streams along the sides of the tank when working at high temperatures .
The end of the pipe protruding from the tank Was inserted in a cylindrical water jacket B , with ends closed by stuffing boxes similar 346 Profs .
H. T. Barnes and E. G. Coker .
* [ Nov. 4 to those on the tank .
From the end of the tank a glass prolongation C held the mercury thermometer placed and securely fastened centrally in the tube .
The glass tube was cemented into a brass tube , connected with a screw-down valve D , for regulating the flow .
From the valve a pipe led to a two-way switch over E , provided with one pipe to waste and one pipe leading to the double-coned copper measure , shown at F. This measure had been very carefully calibrated by Mr. T. P. Strickland some time previously .
For small flows , where the time for filling the tank became inconveniently long , a smaller vessel was used and the amount of discharge obtained by weighing on a large bullion balance .
For the high temperature experiments , a worm pipe G was inserted in the outflow so as to cool the water before it entered the measure .
Water was circulated continuously through the jacket by means of a centrifugal pump H , operated by a Pelton water motor I. A gas flame J , under the pump , served to regulate the temperature of the jacket .
A thermometer K was placed in the water circuit so that the temperature could be regulated at will and maintained a few degrees above or below the temperature of the water flowing out from the tank .
A graduated disc L was provided on the valve , by means of which a. pointer on the valve gave an indication of the amount opened .
During the greater part of the experiments , the temperature of the tank was taken from a long-stem thermometer placed so that its bulb reached to the centre somewhat below the flare on the pipe .
It was found that the temperature thus indicated differed but little from the stream-line reading on the thermometer in the glass prolongation at high temperatures , and was in practical agreement with it at the lower points .
For the majority of the experiments , an average head of about 8 feet was maintained , sometimes running down to 4 feet , but seldom getting as low as that .
One filling of the copper measure only lowered the head f of an inch , so that no constant level device was necessary .
The temperature of the tank could be raised to any point up to 90 ' C. by means of a steam heater .
A stirrer was also fitted to the tank , but of course was not kept moving after the temperature of the tank was once brought to a uniform state .
The method of making an experiment was very simple .
One observer stood at the valve and regulated the flow until it was seen by the reading that the stream-lines had completely disappeared , and the other observer then switched over the flow into the measure and took periodic readings of the temperatures .
The time was taken on a chronograph during the earlier runs , but was changed to a Frodsham and Keen chronometer which had been carefully rated .
The water in the tank was always allowed to remain undisturbed for several hours before taking an experiment .
For the high temperature experiments the tank was heated to about 90 ' and then allowed to settle .
Over a 904 .
] The Flow of Water through Pipes .
tin n(i ) Y/ // // // y/ // // // // // // // // // // / ' / S/ // 77X Y/ // // /S/ // // // // 7/ A VOL , LXXIV .
Profs .
H. T. Barnes and E. G. Coker .
[ Nov. 4 week was required for the tank to cool down , during which tim successive readings were made .
No two determinations were eve made with one setting of the valve , hut in every case a re-determim tion of the critical velocity was made .
It might be questioned whether the presence of the thermomete bulb in the outflowing stream could have produced a disturbing influencr but in no case was the bulb placed inside the tube under test .
W think that our very high values obtained in many cases for the critic ?
velocity show conclusively that the bulb had no influence , and especial !
in the tubes we used , which were all considerably larger than thj thermometer bulb .
In order to satisfy ourselves on many points in connection with on ) measurements , we made some tests to determine the effect of a chang " of length on the critical velocity .
To do this we first took a pip 3'6 metres long and 1*95 cm .
in diameter , and determined the critic : velocity by moving the temperature jacket to the end where the gla ?
prolongation began .
We obtained the following readings:\#151 ; Mean Temp. 16-0 Yc in ft. per sec. 3-79 Yc in metres per sec. 1-155 16-0 3-82 1-163 16-1 3-66 1-115 16*1 3-92 1-192 16-05 3-80 1-158 The tube was then removed and cut in the middle , so as to be on !
1*8 metres long , and the following readings were obtained:\#151 ; Mean Temp. Yc in ft. per sec. Yc in metres per sec. 15-4 3-66 1-115 16-7 3-84 1-170 16-7 3-82 1-163 16-2 3-77 1-148 The agreement is really better than the individual observation : For each of these readings the flow was closed off and then openei carefully until the jump took place on the thermometer .
The reading could have been repeated to any desired accuracy of course by leavin the flow unchanged .
The formula for calculating the critical velocity which Reynold deduced from his experiments for the upper limit reads : y = H c 43*7 D where D is the diameter of the pipe in metres , and P is a function o ; )Q4 .
] The Flow of Water through Pipes .
'0 temperature , which was found to be the formula obtained by kseuille for the change in flow of water in capillary tubes .
jp then equals ( 1 +0-03368 T +0-000221 T2)-1 .
Calculating the upper limit for this pipe we find : v _ 0-0228 c 0-01.95x1-59 0-734 .
This value comes out 0*734 metre per second instead of 1*15 , as we .served it .
The higher limit obtained in our measurements is idoubtedly due to the steadiness of the water at the mouth of our pipe .
We convinced ourselves in these experiments that we had not ceeded the upper limit , for the valve controlling the flow was moved ?
ry slowly as the point was reached , and stopped instantly the jump ok place .
It was possible to obtain the stream-lines again by slightly osing the valve .
For a velocity of 0-7 metre we had perfectly steady ream-lines .
For our smaller pipe , 1-05 cm .
in diameter , we found a ose agreement with the formula of Reynolds , but that above this , for pes of larger diameters , the deviations became larger the larger the ipe .
If we tried the experiments too soon after filling the tank , before te water had become perfectly steady , the upper limit came much iwer , and , as the water became more settled , a higher limit was cheated by a tendency to form stream-lines , followed by a breaking-p , giving rise to the phenomenon described by Reynolds as " flashing .
" he flashes became less frequent until , finally , the water being erfectly steady , the higher limit was reached .
It was possible to btain the higher limit as soon as the flashes started , since it was aly necessary to increase the flow until the flashing ceased .
The ashes showed themselves by an oscillation on the thread of the lermometer .
\#166 ; It was not , therefore , really necessary to have absolutely quiet water 1 obtain the upper limit , since we were able to satisfy ourselves by Jtperiment that the flashes indicated a disturbance in the tank , which , removed , would enable us to reach the upper point without their ppearance at all .
Hence , flashing was taken to indicate a tendency to )rm stream-lines , and if they appeared , the upper limit was obtained y increasing the flow until the permanent reading on the thermometer idicated that eddy motion was the steady flow .
Stream-Line Flow at High Velocities .
|^The 1-8 metre length of our brass pipe had remained attached to the rank for several days under an 8-foot head , after completing our fflperiments , to test the influence of length .
On returning to this pipe , Profs .
H. T. Barnes and E. G. Coker .
[ Nov. 4 we not only passed our former limit , but were unable to make th stream-lines break up until we produced an artificial disturbance b giving the pipe a sharp rap .
The change in flow produced by the was seen by the jump on the thermometer reading .
It returned to tb stream-line reading as soon as the rapping ceased .
We , unfortunately did not measure the flow in this case , but we at least doubled tb previous value of PI48 metres per second .
Being somewhat surprise by this result , we made some experiments on this same pipe by colon ; bands , instead of the thermometer , in order to satisfy ourselves tbs the high values we obtained were not due to some peculiarity of thj thermal method .
After some trouble we introduced the colour tulj into the tank , and dispensed with the heating jacket .
For this w !
were obliged to let the water out of the tank , and introduce free water .
Owing to the disturbance thus caused , it was several daj before we could proceed with the experiments .
It was a simple matter than we had supposed to observe the colour band , since we ha removed the thermometer from the glass prolongation , the thread \lt ; \#166 ; colour could be distinctly seen issuing from the brass pipe .
Hfl obtained , in these experiments , a critical velocity at the same point | before , but , on opening the valve , the stream-lines re-formed at highe velocities , and persisted to the highest velocity we could produce .
Th corresponded with our thermal results , except that , in this case , we ha an actual re-formation .
of stream-lines at the higher velocities afte eddies had appeared at the usual point .
In order to study the re-formation of the stream-lines more easily a glass pipe , specially made for critical velocity experiments , flare blown on to the tube , was inserted , and a colour tube was put i= as before .
The tube was 1*2 metres long , and 1*47 cms .
in diamete No thermometer was used , or heating jacket , and the head was aboi 8 feet , as before .
We found , incidentally , that by increasing tl quantity of colour it was possible to cause a very large disturbance the flow .
The following readings were obtained :\#151 ; Vc in feet Temp. per sec. Little colour ... ... 20 ' C. 3 * 466 n " 20 3-366 Excess of colour ... .
20 1'252 Little colour ... ... 19 3-786 Vc in metres per sec. 1-056 1*026 0- 382 1- 154 These observations showed the importance of using little colour , | a small colour stream produced practically no effect .
Working out the value of Yc as before from the formula , we g 20 ' C. the value 0"88 metre per second .
This value is , again , m. less than the observed value using little colour .
:)04 .
] T^e Flvwof Water through Pipes .
We observed that by opening the valve , and increasing the flow , the ream-lines appeared to re-form .
This was shown by a return of the [ read of colour .
It was not until we reached a velocity of 2*97 metres L second that the thread again disappeared .
By altering the flow a itle at this point we could make the thread of colour disappear , or itain it clearly defined .
We repeated this several times , and found tat the definite nature of this point was remarkable , the thread of dour appearing in almost as sudden a manner as it disappeared , j The diameter of the pipe plays an important part in obtaining the igher stream-line flow , for we found that with a pipe 1'05 cms .
in lameter we could not pass the upper limit , nor cause a re-formation of ream-lines .
We made some experiments with a brass-pipe , 5*41 cms .
i diameter , and 1*5 metres long , by the method of colour bands , and imnd that we could carry the stream-line flow up to velocities of j metre per second , which was the highest flow we could measure .
, 'o obtain this flow we were obliged to arrange a much larger measure y handle the water discharged .
The upper limit for a pipe of this ze , according to the formula , amounts to about 024 metre per second t 20 ' C. We , therefore , exceeded this by four times , as far as we ould see , without the formation of eddies .
There was a tendency to ksh at the highest point , but no definite critical velocity , and the bread of colour could be seen very distinctly .
We cannot enter into a discussion of the influence of the diameter of he pipe on the attainment of the second stream-line flow , but it appears to us obvious that we were able to obtain these higher readings my by paying the strictest attention to the steadiness of the water in i he tank .
The magnitude of our tank , and the volume of water at our lisposal , made this comparatively easy .
The inverse diameter law has 0 \gt ; een shown by Reynolds to be true both for the upper and lower \lt ; imits , and our experiments show the same where the water has not oecome perfectly steady ; but it is probable that , as the diameter of the oipe becomes larger , the disturbing influence of the walls becomes less jffective in causing a breaking up of the stream-lines .
In the jet experiments , where there is no directing pipe , stream-line is the stable low for all velocities , providing the water has become absolutely quiet .
For pipes of small diameter , under ^ inch , or 1 cm .
, the steadiness of the water probably becomes less important , compared with the influence f\gt ; % of the walls of the pipe .
It is probable that for absolutely quiet water If | the inverse diameter law holds up to 1 cm .
, beyond which the critical velocity apparently increases with increasing diameter , until for large pipes we approach the jet .
The higher critical velocity may be a second critical velocity , but we have not decided this point .
Profs .
H. T. Barnes and E. G. Coker .
[ Nov. 4 Temperature Effect on the Upper Limit by the Thermal Method .
In Table I we give the series of experiment obtained for the effect o temperature on the upper limit for our smallest brass pipe , 1-05 am in diameter .
The results are worked out for each point , correcting th volume of the water in the measure to the volume at the temperatur of efflux from the tank .
The velocity of efflux was then calculated ii the usual manner.- In fig. 2 we give a plot of the observations , which extended from 15\#151 ; 86 ' .
Unfortunately , we were unable to extents below 15 ' , on account of the temperature of the water in the mains no !
being below that at the outset of the experiments .
During the time o| the experiments , June and July , after the beginning of the hot weather the temperature steadily rose in the mains to as high as 20 ' , where i stayed during the remainder of the work .
Table I. D = 0*0105 metre .
Total quantity Total time Yc in metres Temperature .
in cubic inches .
in seconds .
per second .
20'*2 3910-7 644-2 1-146 40-1 3898-7 1085-4 0-677 34*1 3891-8 929-9 0-789 29*1 3894-5 811-4 0-905 23-6 3897-9 675-8 1-088 22-8 3895-6 674-3 1-091 56-6 3918-7 1753-0 0-421 49-1 3908-8 1592-7 0-466 84-3 1781-8 245-1 0-137 72-9 .
1778-1 129-6 0-259 47-3 3893-8 1442-0 0-512 23-3 3899-4 675-5 1-091 20-2 3903-8 643-8 1-149 15-09 3902-9 525-7 .
1-401 In Table II we give Reynolds ' observations for his two smaller pipes the diameters of which lie just above and below our pipe .
These tests were made with colour bands .
In the last column we reduce th* observations to a pipe ' equal in area to ours , by the simple ratio of th* diameters .
In the plot we have included Reynolds ' observations , an*j indicated them by circled crosses .
We have , also , included tW temperature law of Poiseuille .
904.1 The Flow of Water through Pipes .
Table II.\#151 ; Reynolds ' Observations on Pipes 2 and 3 reduced to a Pipe 1*05 cm .
in diameter .
Yc in metres Yc in metres Temperature .
per sec. Temperature .
per sec. 22 ' 1-086 6 ' 1-849 11 1-478 - 6 1-869 11 1-489 6 1-838 11 1-505 6 .
1-957 11 1-528 6 1-978 11 1-556 4 1-981 4 1-891 4 2-027 | In the next table we give a summary of the observations which we bbtain for our larger brass pipe at various temperatures .
Table III .
Temperature .
Y in feet , per sec. Y in metres per sec. 21'-3 3-458 1-054 34-5 2-059 0-628 40-0 1-791 0-546 510 1-362 0-415 55-0 1-115 0-340 70-8 0-631 0-192 73-6 0-777 0-237 These results are plotted in Fig. 2 and represented by circled dots .
The general slope of the two curves is the same and , although the agreement between the different determinations is not perfect , it is sufficient to show that the upper limit falls off more rapidly than the theoretical law .
The temperature variation of the upper limit may , from these observations , be represented by the formula P = ( 1 + 0-0300 T + 0-000704 T2 ) .
We do not think that the divergence indicates a temperature variation for the critical velocity different to the theoretical , but rather that it shows , as we have pointed out in a previous part of this paper , that the true critical velocity is at the lower limit .
The inverse diameter law .does not hold for our larger pipe , as will be seen by reference to the plot , and we have shown that it does not hold for the upper limit in the case of our other large pipes .
Temperature Variation of the Lower In a note by one of the authors ( H. T. B. ) , read before the Belfast meeting of the British Association , it was announced that the thermal * Compare also E. G. Coker and S. B. Clement , loc. cit. Profs .
H. T. Barnes and E. G- .
Coker .
[ Nov. 4 , method had been applied to the measurement of the lower limit of stream-line flow , and that , from measurements at different tempera tures , it had been found that the variation with temperature followed the theoretical accurately .
The thermal experiments were arranged in the following way A large copper tank was fitted with a lead pipe , 8 feet long , coiled up in a smooth spiral .
One end of the pipe was connected with the water mains , and the other protruded from the side of the tank and passed through a water-bath , 2 feet long , the temperature of which could be regulated at will and maintained constant by a special form of continuous electrical regulator .
The pipe was fitted with a glass prolongation , similar to our previous experiments , and#a thermometer was used as before .
The tank was fitted with an electrical heater , by means of which the temperature could be changed quickly at will and maintained constant .
For the low temperature experiments , an ice-and-water mixture was placed in the tank .
; 1904 ] The Flow of Water through Pipes .
It was found that it was possible to measure three flows : that at which the flow was entirely stream-line with no tendency to form j eddies ; second , that at which the eddies remained in the flow without the appearance of stream-lines ; and third , that at which the eddies and stream-lines followed each other at regular intervals .
The change from one flow to the other was observed by the rate of the oscillation on the thread of the thermometer .
An inspection of the curves given by Reynolds for the relation between velocity and pressure , above and below the critical velocity , shows that no sharp line of intersection exists between the curve representing stream-line and that representing eddy motion , but that there is a portion over a considerable range of flow where the readings are unsteady .
It was in this region that the three flows mentioned above were found .
The first was the highest limit of the stable stream-line , the second the lowest limit of the stable eddy flow , and the third the point half-way between .
The results are given in the following table for the third point , which is the critical velocity of Reynolds:\#151 ; Table IY .
Diameter of pipe , 0*0125 metre .
Q. Time .
Temp. C.C. secs .
Vc P. UL .
6 ' 830 30*2 0*225 0*238 0*239 6 866 30*2 0*234 0*238 0*239 17*2 952 45*2 ' 0*171 0*176 0*167 17*2 930 45*0 18*1 951 45*1 18*1 926 45*2 18*1 944 45*1 0*171 0*172 0*163 18*1 960 45*0 18*1 923 45*2 40*5 580 45*3 40*5 .
568 45*2 0*103 0*104 0*086 40*5 565 45*1 The first column contains the temperature at which the critical velocity was measured .
The second column contains the total quantity of liquid Q which was collected in the measuring glass during the time given in the third column .
The times were all taken on a stop-watch .
Under Yc are given the values of the critical velocity , calculated in the usual way from the area of the pipe and the flow per second .
Under P the values of the critical velocity are given , calculated by Reynolds ' lower-limit formula .
This formula reads\#151 ; Yc = 1 P 278 *D in metres per second , where P and D have the same meaning as before .
356 Dr. H. J. S. Sand .
The R61e of Diffusion during [ Nov. 22 In the last column , UL , the values of Vc are given , calculated by the same formula , but using for P the temperature variation deduced from the experiments on the upper limit .
The correspondence of columns 4 and 5 , and the divergence in column 6 , are quite sufficient to show that the temperature variation of the lower limit is in agreement with the law of viscosity rather than with the formula obtained from the upper limit .
The experiments are important in showing a close agreement with the theoretical formula of Reynolds , which requires not only the viscosity formula to hold , but the inverse-diameter law as well .
The agreement of the individual observations confirms our previous conclusions , in showing the uncertainty in measuring critical velocity , not from the want of accuracy in the measuring appliances , but from the variation in the point itself .
It is only by taking a mean of a number of observations that anything like an accurate value can be obtained .
" The Bole of Diffusion during Catalysis by Colloidal Metals and Similar Substances .
" By Henry J. S. Sand , Ph. D. , M.Sc .
, University College , Nottingham .
Communicated by Professor J* H. Pointing , F.R.S. Received November 22 , \#151 ; Read December 8 , 1904 .
In a paper on reaction-velocities in heterogeneous systems , Nernst* has recently put forward the view that all chemical reactions taking place on the boundary of two phases proceed to equilibrium practically instantaneously , and that the velocities actually observed are simply those with which diffusion and convection renew the reacting material at the boundary .
As a special instance of heterogeneous reactions , he mentions catalytic decompositions due to finely divided particles , such as colloidal metals , and he believes it probable that the kinetics of these reactions can be deduced from the assumption that chemical equilibrium remains permanently established on the surface of the particles .
Processes of this kind have recently undergone a considerable amount of experimental investigation , the most important instances being the catalytic decomposition of hydrogen peroxide due to colloidal platinum !
and colloidal goldf by Bredig and his pupils , and the catalytic decomposition of the same substance due to hsemase by Senter .
S In his last paper Senter discusses his reaction in the light of Nernst 's hypothesis , and arrives at the conclusion that the known facts do not contradict it .
* * Ztschft .
Phys. Chem./ vol. 47 , p. 52 ( 1904 ) .
f Bredig and Muller , v. Berneck , c Ztschft .
Phys. Chem , , ' vol. 31 , p. 258 ( 1899 ) $ Bredig and Ikeda , ibid. , vol , 37 , p. J ( 1901 ) .
X Bredig and Reinders , 1 Ztschft .
Phys. Chem. , ' vol. 37 , p. 323 ( 1901 ) .
S Ibid. , vol. 44 , p. 257 ( 1903 ) ; ' Roy .
Soc. Proc./ vol. 74 , p* 201 ( 1904 ) .

