]\gt ; diation in a Gaseous ) losiou .
By BERTRAM HOPKINSON , ( Received IsIay 24 , \mdash ; Read June 9 , 1910 .
) In the first report of the British Association ittee on Gaseous Explosions , attention was drawn to the probable importance of radiation in determining the rate of cooling of the mass of hot gas produced by an iuflammable mixture in a closed vessel .
In the second report reference was made to some experiments which I had made on the effect of the walls of the explosion vessel with bright tin-foil .
It was found that if a mixture of coal-gas and air of given composition were exploded in a vessel thus lined , the maximum pressure reached was nearly the same as that given by an identical mixture when the tin-foil was blackened , but the rate of cooling was decidedly less .
Though the experiment established a substantial difference between the rates of heat-loss in the two cases , it was hardly sufficient to justify giving quantitative results , nor was it absolutely conclusive as to the cause of the difference , though it seemed highly probable that it was due to the difference in power of the walls to absorb radiant energy .
I have recently completed a more elaborate series of tests , which I think are sufficiently accurate and conclusive for publication .
Instead of coating the inside of the vessel with tin-foil , I have had it plated with silver , which has in successive explosions of the same mixture been first highly polished and then blackened .
As in the vessel lined with tin-foil , the cooling is slower when the walls reflecting , but the difference is greater with the silver-lined vessel , to about 50 per cent. during the first quarter of a second .
Moreover , there is a difference in maximum pressure , to about 3 per cent. Becords have also been obtained of the change of temperature , during the explosion and subsequent cooling , of a bolometer placed outside a window of fluorite in the wall of the vessel , and these have shown that for one-tenth of a second after maximum pressure something like one-third of the whole heat-loss to black walls is due to radiation , and that the radiation is still perceptible half a second after maximum pressure , when the temperature of the gas has fallen to about 1200o C. Of the whole of the heat given up by the gas in cooling from 2100o C. to atmospheric temperature at least one-fifth part goes in radiation .
The explosion vessel employed was cylindrical , the internal dimensions being 30 cm .
by 30 cm .
It was made of cast iron , and the inner surface Prof. B. Hopkinson .
[ May 24 , after finishing in the lathe was electro-plated .
The polishing was done at first by hand with a leather and rouge , but though this gave a surface which to all appearances was perfectly polished , it was found that appreciably higher pressures and much slower cooling could be obtained by putting more work into the polishing .
Accordingly in the later trials a motor-driven buffing wheel was used , and the surface was also carefully washed with alcohol .
For blackening the surface , a dead black paint consisting of lampblack mixed with a little shellac was put on in a thin layer .
The weight of this layer when dry was about milligramme per square centinletre , and the thickness about mm. The explosive mixture consisted always of 15 parts of idge c having a higher calorific value ( as determined a Boys ' calorimeter ) of about 370 rade thermal units per cubic foot , or say 6000 calories per litre under standard conditions , to 85 of air ; this being very nearly the mixture of maximum strength consistent with complete combustion .
The mixtule was at atmospheric pressure and temperature before .
The products of combustion contained ( by volume ) about per cent. of , and 20 per cent. of , the remainder being with a small amount of excess O. In order to secure strictly comparative results a mixture of gas and air containing about 19 per cent. coal gas was stored in a separate tank for from 3 to 12 hours .
A portion of this mixture was then transferred to the partially exhausted explosion vessel , being there diluted with sufficient air to give the required of 15 per cent. There was sufficient mixture in the tank to give three or four explosions , and , the quantity of air left in the explosion vessel before ad1nitting the mixture from the tank only about one-fifth of the whole , small errors in the measurement of this quantity were without appreciable effed upon the strength of mixture , which must have been substantially identical for all charges taken from the same filling of the tank .
The first explosion was taken with the vessel highly polished .
It was quickly followed with a second explosion of the same mixture , the walls having been in the meantime either repolished or blackened , and a third and sometimes a fourth explosion with blackened walls was taken .
In this way three or four explosions were made , one or more of which were with polished walls and the remainder with blackened walls .
The conditio1ls for these explosions were identical not only as regards the composition of the charge but also in respect of barometric pressure and temperature .
The pressures were recorded in about 20 cases by a pencil indicabor of the outside-spring type made by Dreyer and Company , the ordinary reciprocating drum of which was replaced by a revolving drum driven 1910 .
] On Radiation in a Gaseous L'xplosion .
157 through worm gearing by an electric motor .
In order to reduce friction , smoked paper and a fine point on the indicator were used instead of the usual pencil .
A facsimile of a so obtained is shown in fig. 1 , from FIG. 1.\mdash ; Record No. 19 .
of original size .
Scale of reproduction , 1 cm .
lbs. per sq .
in .
which it will be seen that there is slight oscillation due to the inertia .
In the results recorded below a correction has been applied for these oscillations , a smooth curve being drawn in each case , but the maximum pressure is uncertain , perhaps by as much as 1 lb. per square inch .
Six diagrams were also taken with the optical indicator of the piston and spring type which I some years ago for gas-engine work .
Two diagrams given by this indicator are reproduced in fig. This diagram shows no Seconds FIG. 2 .
oscillation , and is probably in every respect more accurate than that given by the pencil instrument .
It may be observed , howeve1\ldquo ; that the errors of the latter instrument would not affect comparisons of different explosions taken under the same circumstances , except perhaps as ards maximum pressure .
In order to show the degree of accuracy reached it has been desirable to give all the results of the 27 satisfactory records which were obtained .
Particulars of the procedure adopted in polishing are also given * These diagrams , as also figs. 4 and 7 , are exact tracings from photographic enl ments of the originals , the tracing beiog subsequently reduced in reproduction for ixturel 9 in escribedOn Rxplosio ) pressure temperature varied from one group to another , and in order to make the results comparable the pressures recorded have all been reduced to a standard barometer of 760 mm. and a standard temperature of C. , ou the assumption that the absolute pressure is proportional to the barometric height and inversely proportional to the absolute temperature of the mixture before explosion .
In the results of one day with another day it should , however , be remembered that the calorific value and composition of the gas vary slightly from day to first at the records of the explosions in which the vere blackened , it will be seen that the two curves from the same mixture at all points within 1 lb. per square inch .
With the optical indicator and ) the agreement is closer , and here it extends also to the pressure .
On account oscillation the pencil indicator is so iable at maximum pressure .
The results may be summarised as follows:\mdash ; Table .\mdash ; Blackened Walls .
Maximum pressure .
Mean .
Mean .
Optical ( 8 cards ) lll6 Pencil ( 8 cards ) ll The optical indicator reads appreciably than the pencil instrument .
is may be partly due to a in calorific value , but it is probable that there is also some difference between the instruments .
Both were calibrated with dead The " " polished\ldquo ; records do not show such reement , and it is at once apparent that the procedure followed in polishing is important , differences hardly apparent to the eye greatly the rate of cooling .
The 13 pencil records with polished walls show at second pressures ing from to .
per square a of 16 lbs. Explosions of identical mixtures give curves which differ at this period by 5 or 6 lbs. per square inch .
This uncertainty disappeared fely but not entirely after record 14 , when the practice was adopted of the surface with methylated spirits so as to remove all traces of the shellac used in the previous .
By , this washing with the use of a motor-driven buff more consistent results were obtained .
On in a Gaseous Explosion .
161 the same mixture , superposed .
The .
table shows the time ( from the moment of ignition ) taken by the to cool to 1700o C. and 1200o respectively with different surfaces .
Since the heat lost by the products of explosion in cooling to a given temperature must be very nearly the :S whatever the late of the walls , the ti1nes are inversely proporCional to the mean rate of cooling .
This ratio is shown in each case , the rate of cooling with the black walls being taken as the unit .
Table III .
It will be seen that the rate of loss of heat to the polished surface varies in different observations from rather over half to about three-quarters of the loss to black walls .
The differences between one experiment and another cannot be ascribed in any measure either to errors of ation or to the indicators .
Indicator errors would affect both records equally and would not affect the ratios of times .
It seems quite certain that in the optical records , which were taken ether after all the others , the surface was not so good .
Further there seems to be no reason to doubt that in explosion 14 the rate of cooling with the polished surface was not much more than half of that with the blackened walls .
The tangents to the two cooling curves at corresponding points ( that is points for which the temperatures are the same ) tend to become parallel as the temperature falls .
The slopes of these tangents are the ratio of the rates of heat-loss , which therefore tend to equality as the gas cools .
Comparisons similar to the above with a tin-foil lining first bright and then black showed that the ratio of rates of cooling down to 1200o C. was about .
lf the difference between this ratio and unity be taken as a measure of the radiation reflected it will be seen that tin-foil reflects for this purpose less than half as well as silver .
It is not surprising , therefore , that no difference in maximum pressure could be detected with the tin-foil lining , .
a pencil indicator .
type Ttandard pattern mGaseous Fecordi aanometer uhese ments wallistic Messrs. W. Paul , but the usual suspension , which would glve a period of about 4 or 5 seconds , was replaced by a much stiffer s.uspension , giving a period of about second ; and for the usual permanent magnet was tuted an magnetised nearly to saturation , which gave a strong and fairly constant field , whose amount could be accurately adjusted to a definite value by a comparatively rough adjustment of the current .
The resistance of was about ohms , and its deflection was recorded photographically on the same drum as that on the pressures were recorded by the optical indicator .
It will be obvious that with the above arrangement the deflection of the alvanometer is nearly proportional to the increase of resistauce of the bolometer strip ; and this is so .
even exact balance is not obtained efore the explosion .
Tn order to the relation between these two quantities , the current was reduced to zero*when the bolometer strip was cold , and the ratio was found between the alvanometer deflection under these conditions and the current in the bolometer strip , which latter was , of course , for this purpose reduced to a fraction of the value employed when taking explosion record .
lf be the current under these circumstances corresponding to auy deflection of the alvanometer , then , when there is a current shown on the ammeter , the proportion of increase in the bolometer resistance to the same deflection of the galvano1ncter is , where is the current in the galvanometer .
Since the rise of temperature is directly deduced from the proportion by which the rises , it will be seen that the calibration of the bolometer , that is , the reduction of galvanometer deflection to temperature increase , requires only an accurate of the ratio of the currents Strictly speaking , this should be done by reducing the E.M.F. in the circuit to zero ( short-circuiting the battery ) and not by means of the adjustable resistance ; and similarly the reduction of the current in circuit should be effected by a reduction in the E.M.F. of the battery .
The resistances in the circuit being then all the same in calibration and in use , the relation between galvanometer deflection and change of resistance is given in the manner described , being independent of the small changes in and consequent on the disturbance of the circuit by the increase of the bolometer resistance .
In practice it is simpler to make the resistances in circuits so large that they can be treated as infinite , and to use them for adjustment of current ; but it must not be forgotten that variation in either of these currents has the same efl.ect as an equal percentage variation of the bolometer istance .
If , for example , it be desired to measure a rise of temperature of C. correct to 1 per cent. , the rise resistance amounts to perhaps per cent. , and consequently or must be to within 1 part in 600 .
1910 .
] On a xplosion .
In the course of heated by the flame the silver loses a considerable quantity of heat to the backing with which it is in contact .
The percentage of heat so lost depends upon the thermal conductivity and the thermal capacity of the backing , and on the manner in which the temperature of the silver rises .
If the two curves for polished and blackened silver were precisely similar , only in respect of the scale of temperature , the proportion of heat lost at corresponding times would also be the same for , and the quantity of heat absorbed would be proportional to the temperature .
The curves are not quite of the same shape , and the small correction applicable on this account has been calculated and is allowed for in the last column .
The figures there given represent , probably within 5 per cent. , the proportion in which the black and polished silver surfaces absorb when exposed to gas temperatures varying in the manner shown by the above record .
It appears that the mean rate of absorption during the first quarter of a second by the polished surface is of the absorption by the lackened surface .
The records of pressure obtained with polished and blackened walls , to which reference is made in section ( b ) above , showed that the mean rate of cooling in a polished vessel during the first quarter of a second was in different experiments from to of the rate observed when the walls are blackened .
The agreement is as good as can be expected , having regard to the great effect of small variations in the state of pohsh .
The bolometer also shows that approximation of the rates of cooling in the two cases as the temperature falls which appeared in the pressure records .
The difference in heat absorption represents no doubt the reater part of the radiant heat emitted by the gas in the time , but it does not represent the whole of it .
For the black surface is not perfectly absorbent , nor is the polished surface perfectly reflecting .
Thus it is quite certain that of the heat given by the gas to the walls of a blackened enclosure duriIlg the first quarter-second after maximum pressure , at least 30 per cent. is radiant heat , and the proportion is probably a good deal higher .
VOL LXXXIV.\mdash ; A. 1910 .
] On in .
167 Applying this method in the present instance , it has been found that : 0.5 second after , when the temperature of the is , the total heat-loss to the walls ( averaged over the whole surface ) is calories per square centimetre .
* The temperature of the ckened silver is then C. , equivalent to calorie per square centimetre in the silver .
The heat lost to the backing at this epoch is therefore calorie per square centimetre .
It is here assumed that the heat absorbed the bolometer is equal to the absorption over the whole surface .
) , we have , therefore , The value of the integral is obtained from the record of temperature ( see fig. 4 ) , and is found to be when .
It follows that A direct measurement of the thermal conductivity and thermal capacity of the linoleum , by methods which need not be described here , gave whence or only abouI half the value given by the other method .
Assunning that is equal to , the following table shows the absolute quantities of heat absorbed by the blackened and polished surfaces per unit area at different times .
The second column gives the temperatures reached by the two bolometers ; the third , the corresponding heat quantities in the silver ( calculated on the assumption that the specific heat of pure silver is ; the fourth , the factor , which gives the proportion of heat lost to ; and the fifth , the actual absorption after this loss .
The sixth column is the difference in the heat absorption of the blackened and polished surfaces .
If the absorption by the bolometer is equal to the over the whole surface of the vessel , the total heat lost to the walls of the blackened vessel at maximum pressure ( second ) is 1080 calories , or per cent. of the whole heat of combustion .
This yrees well with the results of experiments See Appendix .
1910 .
] On in Explosion .
connections of the bolometer were the same as in the experiments last described ( blackened and polished silver\mdash ; see fig. 3 ) .
The bolometer , which was blackened with camphor smoke , was pushed into a gun-metal tube screwed into a boss on the end-cover of the vessel , which carried at its inner end the plate of fluorite ( fig. 6 ) .
FIG. 6 .
The fluorite was -inch thick , and the clear inch diameter .
The interior of the vessel was blackened .
One record taken under these conditions during the explosion of cent. mixture is reproduced in fig. 7 .
The figure is an exact tracing from an enlargement of the original .
Ib is noteworthy that there is no oscillation of Prof B. Hopkinson .
[ May 24 , the galvanometer , doubtless because there is no sudden heating such as occurs when the flame touches a surface in contact with the gas .
The following table shows the results of this record:\mdash ; Table The loss of heat from the platinum by conduction and radiation ( shown in the fifth column ) was determined by calculation from the rate of fall of temperature in the later stages of cooling , when no radiation was being received .
The percentages in the last column are calculated on the assumption that radiant heat is absorbed by every part of the blackened walls at the same rate as by the bolometer , no allowance being made for absorption in the fluorite or reflection from the bolometer .
Probably about 10 per cent. should be added on this account to get the true amount of radiation .
The amount of radiation shown in this experiment is from 40 to 50 per cent. in excess of that given by comparison of the blackened and polished silver bolometer inside the vessel , at corresponding times .
The uncertainty attaching to the latter observations , partly because they are difference measurements , and partly because of the very large corrections for loss to the backing , will not have escaped observation .
I think , however , that the disclepancy is too great to be accounted for in this way , and that the difference between the absorption by the blackened and polished silver surfaces is really a good deal less than the whole of amount of radiation . .
W. T. David has suggested that a film of water may perhaps be formed on the polished surface at an early stage of the cooling andimpair its reflecting qualities .
About 5 grammes of water are produced in the explosion , and practically the whole of this ultimately condenses on the walls , covering them to an average depth of of a millimetre all over , thus destroying in large measure their reflecting power .
It is probable that the formation of this layer commences almost as soon as the flame touches the metal , and that a much thinner layer uorite pther hmust amost iecomeWOlll 9dequate tbsorb araction oadiation falling too hot on the surface to admit of the deposit of moisture .
It is possible that the rate of formation of this film depends on the exact state of the metal surface , and that we have here the explanation of the great effect of apparently inappreciable differences in polish on the rate of cooling .
found that if a plate of glass was substituted for the fluorite , the heat received by the bolometer was reduced to about one-third ; and that if the platinum were polished the heat received was only about 20 per cent. of that recorded with a blackened surface .
The latter proportion agrees with the figure given by Hagen and Rubens for the reflecting power of polished platinum .
The experimental work described in this paper has been carried out by Mr. W. T. David , advanced student at the Engineering Laboratory , to whom I am much indebted for the ability and zeal with which he has prosecuted the research and for many suggestions of value .
He is engaged upon experiments with the recording bolometer similar to those last described , by which it is hoped to determine the nature and origin of the radiation , and its dependence on temperature and density .
The work of R. von Helmholtz*and of provei that the radiation from a nonluminous flame ( .
a Bunsen flame ) at atmospheric pressure consists mainly of two bands , of wave-lengths and , which come from the carbonic acid and steam respectively .
It is probable that the radiation in a closed-vessel explosion is of substantially the same quality , and Mr. David 's work , so far as it has gone , confirms this view .
lation of Hjat-Loss from Pressure Record .
At second fro1n the time of ignition the mean temperature of the gas calculated from the pressure is nearly C. The energy is shown in the } table:\mdash ; ' Die Licht- und Warmestrthlung verbrennender Gase , ' Berlin , 1890 .
Die Licht- und irmestrnhlung verbrannter Gase , ' Berlin , 1890 .
It has been suggested that possibly some of the radiation comes from the walls of the vessel , which are heated the explosion .
From consideration of the rate of flow of heat per sq .
cm .
, however , it be shown that the surface temperature of the metal can never exceed about 40 ' C. , which is of course quite inadequate to give any appreciable radiation .
Further , Mr. David has found that if a patch cf the wall opposite the bolometer be highly polished , the radiation received is .
The fluorite plate will be heated to a fairly high temperature at the inner surface , but the radiation emitted by this surface will be absorbed in the fluorite and none of it will reach the bolometer .
No appreciable heating of the outer surface can ocour during the measurement of the radiation .

