]\gt ; PHILOSOPHICAL TRANSACTIONS .
I. BAKERIAN LECTURE.\mdash ; On the Variation of the Specific Heat of Water , with Experiments by a New Method .
H. L. CALLENDAR , M.A. , .D .
, F.R.S. , of Physics at the College of Science Technology , London , S. Received December 5 , 1911 , \mdash ; Received in extended form as Bakerian Lecture , February 6 , 1912 , \mdash ; Lecture delivered Febru , 1912 .
THE question of the mode of variation of the specific heat of water is so fundamental in calorimetry , there are so few experiments between C. and C. , and the results in this region obtained by different observers are so discordant , that no apology is needed for the publication of new experimental work tending to throw light on the subject .
But in order to elucidate the points at issue , it will first be necessary to review briefly the experimental evidence already in existence , and to exhibit the results graphically as an indication of the order of accuracy of the various methods .
A minor difficulty in comparing the results of different observers arises from the fact that they are expressed in terms of different thermometric scales and units , and that the reduction to a uniform standard of comparison cannot always be effected with certainty .
Throughout the present paper , for reasons which have been fully explained elsewhere , all heat quantities are expressed in terms of the specific heat of water at C. taken as unity .
The scale of temperature , , adopted is that deduced from the temperature pt by platinum resistance thermometer by means of the rence formul , . . . . .
( 1 ) Many computers have corrected this formula by assuming values from C. to C. , or even C. , for the sulphur boiling-point on the perfect-gas scale .
But the recent experiments of HOLBORN and HENNING ( Ann. Phys 35 , pp. 761-794 , 1911 ) with a quartz-glass bulb give the value C. on the perfect-gas scale .
They assume the linear expansion coefficient of qual.tz-glass constant and eqnal to , which makes the cubical coefficient 1 .
But it appears that the linear coefficient vanishes at C. , and is likely to be smaller between C. and C. at higher temperatures .
In any case it is unsatisfactory to deduce the cubical coefficieut from the linear , because the latter is difficult to measure ately , and may well be different in differe1lt directions for a drawn bulb , especially as quartz-glass cannot be annealed owing to its rapid devitrification at temperatures in the rhood of 1,000o C. Direct measurements of the cubical coefficient of a quartz-glass bulb , by E. J. HARLOW ( Proc. Phys. Soc Lond. , Nov. , 1911 ) , employing the method of the mercury weight VOL. CCXII .
16.4.12 which was proposed at the British Association meeting in 1899 as an easily reproducible scale of reference for the experimentalist .
The correction of the practical scale , as above defined , to the absolute scale is so small and uncertain , and has so often been applied incorrectly , that its application appears more likely to lead to confusion than to improved agreement in experimental work .
The specific heat of water at C. has often been adopted as the standard .
It exceeds that at .
by little more than 1 in 1,000 .
The reduction from C. to C. can be effected with comparative certainty , but does not materially affect the question of the variation of the specific heat , since all the values are altered nearly in the same proportion , and few results are accurate to 1 in 1,000 .
REGNAULT'S Experiments , C. to C. REGNAULT operated by mixing 10litres of water from a boiler at various temperatures between C. and C. with 100 litres of water in a calorimeter at the atmospheric temperature .
His observations gave directly the mean specific heat*of water from the temperature and pressure of the boiler to the final temperature of the calorimeter at atmospheric pressure in terms of the mean specific heat between the initial and final temperatures of the calorimeter .
The results did not give any direct evidence with regard to the variation of specific heat between C. and 10 C. , but were fairly consistent over the range C. to C. with the assumption of his well-known parabolic formula . . . . .
( 2 ) for the mean specific heat between C. and C. ROWLAND'S discovery , in 1879 , that the specific heat of water diminished by about 1 per cent. between .
and C. , showed that REGNAULT'S formula could not possibly be correct at low temperatures , and necessitated a recalculation of his thermometer , and assuming the absolute expansion of mercury from the observations of CALLENDAR and Moss Phil. Trans. give a value 1 between C. and C. , increasing to 1 between C. and C. If this should be correct , the result of HOLBORN and HENNING would require to be raised by C. , giving C. on the absolute scale .
It is hoped that experiments now in progress by N. EUMORFOPOULOS by the constant-pressure method , with a quartzglass thermometer of the Callendar type , will throw further light on this important point .
It would appear in any case that formula ( 1 ) gives a better approximation to the absolute scale of temperature than has previously been supposed .
It has been possible for the last twenty years to obtain platinum for thermometric purposes of the same unvarying degree of purity , giving a temperature coefficient approximating to , and a practically constant difference-coefficient .
It seems , therefore , preferable to eliminate errors of observation of the sulphur boiling-point , and differences of opinion as to its absolute value , by assuming a standard value for the difference-coefficient in the deflnit.ion of the actical scale .
* More accurately , the change of total heat HEAT OF WATER , WITH EXPERIMENTS BY A NEW IsIETHOD .
experimental results .
This was undertaken by J. M. GRAY ( ' Proc. Inst. Mech. Eng 1899 ) who found that the data by REGNAULT did not in all cases agree with his calculated results , the discrepancies occasionally reaching 2 or 3 per cent. It appeared , from measurements of the original apparatus , that in all these cases the recorded quantity of water exceeded the total capacity of the calorimeter .
GRAY concluded that REGNAULT'S calculations were probably correct , and that the discrepancies arose from deficient information or erroneous entries in the data columns .
Adopting this assumption , it is possible to recalculate ' observations , allowing for the known variation of the specific heat from C. to C. , and to express his results for the mean specific heat from C. to C. in terms of the specific heat at C. The separate observations , reduced in this manner , are represented by the small crosses in fig. 1 .
The crosses surrounded by circles Fig. 1 .
indicate the means of each group .
The general effect of this reduction is to bring the results on the average about 2 parts in 1,000 below REGNAULT'S formula .
The correction is fairly certain , and is less than the discrepancies between the observations in any one group .
The probable errors of REGNAULT'S thermometers remain to be considered .
These have not been included in the reduction , as being much less certain , because the original thermometers cannot now be recovered and tested .
It appears recent observations on the absolute expansion of mercury , and on the of steam between C. and C. that REGNAULT'S temperature scale over this range did not differ materially from that given by formula ( 1 ) .
It is certain , however , that REGNAULT was ignorant of the phenomenon of the temporary of zero of a mercury thermometer when heated , and that he was unable to detect any systematic difference between the mercury and gas scales of temperature over the range C. to C. , both of which facts would appreciably affect the reduction of the readings of the thermometers employed for observing the rise of temperature in the calorimeter .
The correction for these two sources of error combined might reduce REGNAULT'S values by 5 or 6 parts in 1,000 if the thermometers he employed were of French " " \ldquo ; glass .
In any case the reduction could hardly be less than 2 or 3 parts in 1,000 .
This would bring the observations into fairly good agreement with my extrapolated formula , indicated by the lower curve in fig. 1 , but the corrections involved are so hypothetical that no great stress can be laid on them .
The only satisfactory solution is to repeat the observations , for which I have already made such preparations as my scant intervals of leisure will permit .
In the meantime we may regard REGNAULT'S observations as giving , with some degree of probability , the rate of increase of the mean specific heat between C. and C. , although the absolute values given by his formula probably require reduction by about per cent. It should be observed that , even if all the corrections could be applied with certainty , the order of accuracy of his final results could not be expected to exceed 1 or 2 parts in 1,000 , because the calorimetric thermometers were read to C. only on a rise of temperature of C. to C. , and the individual observations in each group show corresponding discrepancies from the means .
REGNAULT himself did not claim any higher order of accuracy , and endeavoured to indicate this by the values of the coefficients given in his formula .
Range C. to C. ( LUDIN ) .
Many of the investigations by able experimentalists extending over the range C. to C. have given rates of variation exceeding 10 per cent. per 10 C. , which were doubtless due to defective experimental methods and insufficient appreciation of the real difficulties of the problem .
Such results are of no value except as an indication that the problem is not quite so simple as it appears at first sight .
The first investigation in which sufficient attention was given to the well-known difficulties of mercurial thermometry , was that of E. LUDIN ( ' Die Abhangigkeit der specifischen Warm des Wassers von der Temperatur , ' Inaug .
Diss .
Zurich , 1895 ) , carried out by the method of mixtures under the direction of Prof. PERNET .
His observations gave directly the mean specific heat over eight different ranges of temperature above C. , and two different ranges below C. , in terms of the mean specific heat over the range C. to C. The variation of the actual specific heat was deduced by assuming a formula of the type , . . . . .
( 3 ) and calculating the values of the coefficients by the method of least squares to agree with the observed ratios of the mean specific heats over the various ranges .
This method is somewhat indirect , and makes the result depend to some extent on the HEAT OF WATER , WITH EXPERIMENTS BY A NEW fHOD .
particular type of empirical formula chosen .
The formula given by LUDIN for the actual specific heat was as ' . . .
( 4 ) The probable errors of the several coefficients , as calculated by LUDIN , are given in the second line below the coefficients to which they apply .
It would appear that this type of formula is unsuitable for representing the variation of specific heat of water from C. to C. , because the coefficients come out relatively large and of opposite signs .
For instance , the value of the specific heat at C. is made up as follows , according to LUDIN'S formula:\mdash ; Sum ?
\mdash ; The small difference , representing the required variation of the specific heat , is less than 1 per cent. of the sum ( taken without regard to sign ) of the Fig. 2 .
terms by which it is represented , and is only a tenth of the probable error of either of the last two terms .
It is obvious a priori that a measurement of the mean specific heat between C. and C. , combined with a measurement between C. and * LUDIN'S results have since been corrected by reference to a comparison made by THIESEN , and SELL between the French ' Verre \ldquo ; and the Jena glass thermometers .
This reduction is somewhat uncertain , and does not affect his values ially except in the neighbourhood of C. His original formula has been retained for purposes of discussion , because it is more nearly correct between C. and C. , and because it was employed by Messrs. BOUSFIELD in their comparisons .
There is an obvious misprint in LUDIN'S corrected formula as quoted in the 'Fortschritte der Physik , ' 1900 , II .
, p. 304 , but the tables appear to be correct .
LUDIN'S ratio of the mean specific heat from C. to C. to the specific heat at C. is reduced from to .
His value for the specific heat at C. is reduced from to in terms of the specific heat at C. C. , coul not give very certain information with regard to the actual specific heat between C. and C. The observations themselves , giving the ratios of the mean specific heats over different ranges , cannot be represented graphically in relation to the curve of actual specific heat represented by formula ( 4 ) , which is shown by the dotted line in fig. 3 .
It is easy , however , to deduce the corresponding formula for Fig. 3 .
the mean specific heat from C. to C. , shown in fig. 2 , in relation to which the observations may be represented by a slight reduction ( as in the case of RHiGNAULT'S observations ) which does not materially affect their relative errors .
The individual observations at each point are shown by the small crosses .
The large crosses surrounded by circles represent the means of each group .
The full curve , marked LUDIN'S formula , represents the equation ' ( 5 ) which gives the mean specific heat from C. to C. by LUDIN'S formula in terms of the specific heat at C. taken as unity .
The group means are seen to differ by less than 1 in 1,000 from LUDIN'S curve .
It must be remembered , however , that the observations in each group were all taken consecutively under the same conditions with the same calorimeter and thermometers , and would give no indication of possible constant errors .
Thus in LUDIN'S method of operation it was necessary to vary the quantity of hot water introduced into the calorimeter from 460 .
at C. to 102 .
at C. , with a corresponding variation in the initial water content of the calorimeter , and in the temperature of the walls above the water level .
In spite of the great care taken in reading the thermometers and applying all corrections to C. , the results of consecutive observations under the same conditions often differ by 2 or 3 parts in 1,000 , especially near the ends of the range .
This is probably due to the uncertainty of heat loss or gain during the transference of the hot water to the calorimeter , of which no account could be taken .
Such gain or loss would HEAT OF WATER , WITH EXPERIMENTS BY A NEW IsIETHOD .
vary , on the whole , systematically with the temperature of the hot water , and would inevitably lead to constant errors .
It will be observed , on reference to fig. 1 , that formula for the mean specific heat , when extrapolated , gives results agreeing closely with 's observations up to C. But this is really without significance , because REGNAULT'S results , as plotted , are not corrected for the probable errors of his calorimetric thermometers , and would certainly require to be further reduced .
LUDIN'S curve , if extrapolated to C. , would give results about 10 per cent. too low , and is obviously of a type which cannot be trusted for extrapolation .
It is almost inconceivable on any theoretical grounds that the specific heat of water , after reaching a maximum at C. , should then diminish and increase again .
LUDIN'S method , as already explained , could not be trusted to give certain results with regard to the variation of the specific heat near the ends of his range .
The experimental evidence for the drop in the curve near C. is very weak , and , such as it is , may be most readily explained by a slight loss of heat due to evaporation of the nearly boiling water on its way to the calorimeter .
It would appear almost hopeless to obtain reliable results by method of mixtures with an calorimeter .
REGNAULT'S method , employing a nearly closed calorimeter of large volume , permanently connected to the heater by a tube for introducing the water , appears to be the only satisfactory of avoiding the uncertainty of heat loss in transference if the ordinary method of mixtures is employed .
Iethod ( CALLENDAR BARNES ) .
The continuous-electric method , in which a steady current of water at any desired temperature is heated through a small range of temperature by a steady electric current , has the great advantage that it gives directly the actual specific heat over a small range at the desired point in place of the mean specific heat over a range , and appears for this reason peculiarly suited for determining the of the specific heat .
The method has been very fully described and discussed in previous papers ( CALLENDAR , 'Phil .
Trans. , ' , 1902 , vol. 199 , pp. ; BARNES , .
cit. , pp. , but it appears desirable to enumerate briefly its principal features .
The form of the calorimeter , being merely a fine-bore tube about half a metre long with enlargements at either end for the thermometers , gives a very small water equivalent and radiating surface , and permits complete enclosure in a hermetically sealed vacuum-jacket , which reduces the external heat-loss to a mininlul1l .
The vacuum-jacket is surrounded by a water-jacket maintained at a steady ture .
The water current is brought to the same temperature as the jacket before passing the inflow thermometer .
The rise of temperature of the water passing through the tube is obtained by a reading on a pair of differential platinum thermometers , sensitive to C. , and probably in all cases accurate to C. , avoiding nearly all the difficulties of mercurial thermometry .
The electric heating current passes through a conductor of bare platinum wire , extending throughout the length of the fine-flow tube , and connected at either end to massive terminals of negligible resistance in the form of thick copper tubes enclosing the thermometers .
These terminals serve to define accurately the resistance of the heater , and also to equalise the temperature along the length of the bulbs of the thermometers .
The fineness of the flow tube renders any insulation of the conductor unnecessary .
Sufficient stirring is obtained by causing the water to flow spirally round the thermometer bulbs and along the fine tube , or by subdividing the heating conductor .
The heat generated by the friction of the water current in passing through the tube could be accurately measured by observin the difference of pressure between the ends of the tube , but its effect on the results was in all cases less than 1 part in 40,000 .
The electric current of 4 to 6 amperes was maintained steady by a battery of very large cells ( 300 ampere rate of discharge ) because this permitted an order of accuracy of 1 or 2 parts in 100,000 in the electrical readings and appeared preferable to discontinuous hand regulation .
For this and other reasons the potential difference on the heating conductor was not directly balanced ainst an integral number of Clark cells , but was measured on a carefully calibrated potentiometer , together with the potential difference due to the same current passing through a specially designed standard resistance of platinum-silver maintained at a constant temperature in an oil-bath .
As is usual in calorimetric experiments , the accuracy obtainable was limited chiefly by the determination of the heat-loss , which was deduced from experiments in which the electric and water currents were varied in such a way as to maintain the same rise of temperature .
The heat-loss for the same rise of temperature was found to be not quite independent of the flow .
It was also found to vary slightly owing to slight changes in } } vacuum , when the apparatus was maintained for some time at a high temperature .
It was inferred , however , that the uncertainty from all causes combined could not have exceeded 1 in 1,000 even at the limit of the range , and was probably much less at lower temperatures .
The experiments have been criticised chiefly on the ground of the uncertainty of the absolute values of the electrical units ten years ago , but this would not affect the question of the variation of the specific heat with temperature , as there was no question of the of the standards employed .
It is true that it was not at that time possible to secure a direct comparison of the Clark cells with the international standards which have since been established .
In default of this I made , with the assistance of Mr. Kin , an absolute determination of the E.M.F. of the cells at that time in use in the laboratory , by means of an specially designed to read to 1 or 2 parts in 100,000 .
I was confident that the result of this determination , namely , volts for this type of cell at C. , would prove accurate to at least 1 in 10,000 .
This has since been verified independently by WOLFF and WATERS Bull .
American Bureau Standards , , who give the value volts at C. , in terms of modern standards , Clark celk HEAT OF WATER , WITH EXPERIMENTS BY NEW METHOD .
constructed with washed mercurous sulphate according to the specification then adopted .
The idea that the absolute values of the mechanical equivalent deduced from the continuous-electric method were uncertain to the extent of more than 1 in 1,000 , and had recently required correction , appears to have arisen from the fact that some of the results were originally expressed in terms of the conventional , or legal , value , volts at C. , of the Clark cell , adopted twenty years ago .
This value was known at the time to be inaccurate , and was employed merely for the purpose of comparing results with those of other observers who had also assumed the conventional value .
The resistance of the standard platinum-silver coil employed in the current measurements was probably known to 1 in 10,000 in terms of then existing standards , but it has been suggested that some uncertainty existed with regard to the heating effect of the current .
The coil was designed with eight wires in parallel , so that its rise of temperature in a well-stirred oil-bath when carrying 8 amperes ( 4 watts in each wire ) , should not exceed C. , or 1 in 4,000 increase of resistance .
The currents actually employed in the specific-heat determinations gave about watts in each wire .
It is , therefore , unlikely that the increase of resistance could have been greater than 1 in 5,000 , even if allowance is made for slight differences between the experimental coil on which the design was based and the actual standard .
The experiments , made by Profs .
VIRIAMU JONES and W. .
AYRTON , with the Lorenz apparatus , which I ordered for McGill College in 1895 , had shown ( CALLEN .
cit. , p. 71 ) that the Board of Trade Standard Ohm was probably 2 or 3 parts in 10 , larger than C.G.S. Since the absolute values of the Clark cells as well as the platinum-silver resistance were determined with reference to this standard , the absolute values of the mechanical equivalent would require to be raised on this account by 2 or 3 parts in 10,000 .
Correction for the heating effect of the current would require the results to be lowered to nearly the same extent .
Since these corrections were nearly equal and of opposite sign , and since both were so small and uncertain , it did not appear desirable at the time to correct for either .
Formula for the from C. to 20 C. When the preliminary results of the continuous-electric method were first announced at the meeting of the British Association at Dover in 1899 , I suggested two simple formulae to represent the variation empirically , one covering the range of minimum specific heat from C. to C. , and the other , a simple modification of REGNAULT'S , fitting the first at C. and representing REGNAULT'S corrected and reduced results up to 20 C. The ffist formula was subsequently modified by the addition of a small term below C. to represent more accurately the rapid increase of specific heat as the freezing-point was approached .
It is for many purposes inconvenient to have to deal with different formula covering limited ranges , however carefully they may have VOL. CCXII.\mdash ; A. been fitted together .
I have , therefore , devoted some attention to devising a single formula of a suitable type to cover the whole range .
Of the many possible types of formula which might be devised for the purpose , the following has appeared to me , after exhaustive trials of many types from different points of view , to be the simplest and the most generally convenient:\mdash ; . . . .
( 6 ) This formula was obtained by combining a formula with REGNAIJLT'S coefficients to represent the variation from C. to C. , with a formula containing a reciprocal term to represent the rapid fall in the neighbourhood of C. The value of the constant term is adjusted to make when C. , which is the most convenient temperature in practice to use as a standard of reference .
The other terms are all small and positive , and can be calculated with sufficient accuracy for all possible purposes by means of a ch slide-rule , which is far from being the case if a formula of the LUDIN type is employed .
This formula is represented by the line in fig. 3 .
The observations of BARNES , represented by the crosses , have been reduced to a unit at C. , and corrected for the variation of the temperature-gradient in the flow-tube , as explained in my paper ( loc. cit. , p. 129 ) .
The results are plotted in terms of the temperature scale defined by formula ( 1 ) , and are not reduced to the hydrogen scale on account of the smallness and uncertainty of this correction , as previously stated .
It may be observed that the agreement of the observations with the curve would be slightly improved if the mean of the large group of observations near C. had been taken as the basis of reduction in place of the few observations near C. This would have the effect of depressing all the points by , but would not alter the form of the curve .
It happens that the absolute value of the specific heat can be most easily determined by the continuous-electric method in the neighbourhood of C. , which would naturally be selected as the standard temperature if this method were the only one to be considered .
None of the observations deviate from the curve by more than 1 in 1,000 , and only seven by more than 1 in 2,000 .
The agreement is very good considering that the observations were taken with several different calorimeters and thermometers at dates extending over more than a year .
Taking account of all the changes of condition which were made in testing the method , it seems hardly likely that the variation of the specific heat given by the formula ( 6 ) can be in error by so much as 1 in 1,000 even at C. C. to C. ( DIETERICI ) .
The earlier experiments of DIETERICI , in which determined the absolute value of the mean calorie C. to C. by passing a current of ampere to ampere , measured with a silver voltameter , through a resistance of ohms in a Bunsen ice-calorimeter , gave a result joules per , assuming the constant of the calorimeter as mgr .
of mercury HEAT OF WATER , WITH EXPERIMENTS BY A NEW METHOD .
per mean calorie from the mean of the values given by BUNSEN , SCHULLER and WARTHA , and VELTEN .
Taking ROWLAND'S values for the equivalent from to C. , and assuming a linear increase from C. to C. to fit with his own value for the mean calorie , DIETERICI deduced a table ( loc. cit. , p. 441 ) for the variation of the specific heat from C. to C. , which has been frequently quoted and employed for reducing observations .
According to this table , the mean calorie exceeded that at C. by 15 per cent. The specific heat at C. was 3 per cent. greater than the value subsequently found by the continuous-electric method , but appeared to be in fair agreement with older observations .
DIETERICI'S later determinations of the meaIl calorie ( ' Ann. Phys 16 , p. 593 , 1905 ) by a similal method , in which the current was reduced to ampere and the resistance increased to 40 ohms in order to diminish errors due to conduction and generation of heat in the leading wires , gave a result joules per .
C. , exceeding the value given by the continuous-electric metbod by only in 1,000 , which is almost , if not quite , within the possible limits of error of the ice- calorimeter .
Accepting DIETERICI'S value of the specific heat at C. , namely , in terms of the mean calorie , his value for the mechanical equivalent at C. would be , which agrees to 1 in 3,000 with the continuous-electric method .
DIETERICI'S value of the calorie at C. has accordingly been taken in place of the mean calorie in reducing his results for comparison with those of other observers .
It should be remarked , however , that the rate of heat supply in his experiments with the ice-calorimeter was 300 times smaller than in the continuous-electric method , and that , in order to obtain equally good results with the ice-calorimeter , it would be necessary that the uncertainty of the heat-loss should also be 300 times smaller , the probability of which is open to doubt .
DIETERICI also redetermined the constant of the ice-calorimeter by an improved method , employing sealed tubes of quartz-glass to contain the water at C. The value thus found was mgr .
per mean calorie , exceeding the value previously employed by 1 in 300 .
His results by the same method for the mean specific heat from C. to C. , reduced to his calorie at C. as unit , are indicated by the diagonal crosses in figs. 1 and 2 .
Between C. and C. , his results , as shown fig. 2 , agree to 1 in 1,000 with my formula , except for one observation at C. , where , as he admits , an inferior degree of accuracy was to be expected .
For temperatures above C. he represents his results for the mean specific heat from C. to C. in terms of the mean calorie by the formula which gives a minimum at C. , agreeing with his observations .
He points out that the corresponding formula for the specific heat at bein obtained by differentiation , has an inferior degree of accuracy , which he sets at per cent. in the neighbourhood of C. The curve marked DIETERICI , in fig. 3 , represents this nula reduced to a unit at by dividing by the factor representing his value at C. DIHiTERICI employs this formula for the actual specific heat , , in calculating the values given in his table down to a temperature of C. , although it does not strictly apply below C. Below C. his values for the specific heat are calculated in a diffelent way , but agree so closely with my formula that his curve could not be shown separately .
This agreement is very satisfactory , but gives rise to a somewhat sharp change of curvature at C. , which is repeated at C. in his formula for the mean specific heat , and introduces some uncertainty in the interpretation or application of the tabulated results .
Below C. his table of mean specific heat appears to represent his experimental results between C. and C. ( neglecting the discordant observation at C. ) with an almost linear extrapolation which follows my curve very closely from C. down to C. Values of the actual specific heat calculated from this table show a rapid fall from at C. to at C. , and a sudden jump from below up to the value at C. If , on the other hand , his table of actual specific heat is taken as the basis of calculation , the value of the mean specific heat from C. to C. comes out in place of , given in his other table .
The observations themselves do not afford any valid evidence for the existence of these discontinuities , which might prove very troublesome in the practical application of his tables .
single continuous formula , such as ( 6 ) , presents many advantages in this respect , especially for representing observations on the mean specific heat , which ought not to show sudden changes of curvature .
The deduction of the true specific heat at any temperature from the mean specific heat is most uncertain in any case , and the observations cannot be said to support the mimmum at C. , shown in DIETERICI'S curve for the actual specific heat .
The uncertainty in the reduction of the results from C. to C. must affect the whole form of the curve , and even the apparent discrepancy of per cent. at C. , shown in fig. 3 , does not exceed the limits of possible error in the calculation .
Apart from variations in the fundamental constant ( depending possibly on the quality of the ice formed ) , and uncertainty of the correction for creep of zero , which might give rise to accidental errors , the main source of systematic error in DIETERICI'S method would lie in the correction for the water equivalent of the quartz-glass bulbs , and in loss or gain of heat during transference from the heater to the calorimeter .
The water equivalent of each bulb was calculated from its mass by means of a formula for the variation of the specific heat of quartz , which is appropriate if there is no heat-loss in transference .
It appears probable , however , that bulbs of different form and thickness would experience different losses in transference .
In fact , a smaU systematic error of this kind is indicated by the observations themselves , and might produce appreciable errors at the higher temperatures .
HEAT OF WATER , WITH EXPERIMENTS BY A NEW METHOD .
Below C. the uncertainty would probably not exceed 1 in 1,000 , as the thermal capacity of the bulbs employed was only a quarter of that of the contained water .
From C. to C. all the observations , except one at C. , were made with bulbs having a thermal capacity nearly equal to that of the contained water .
In the observations at 10 C. , C. , and C. , where both thick and thin bulbs were employed , the results deduced from the thick bulbs , assuming the same formula for the specific heat of quartz-glass , were systematically higher by , and per cent. respectively than those deduced from the thin bulbs .
The point shown at C. in fig. 1 , which is the lowest depending entirely on observations with the thick bulbs , shows so great an increase , when compared with observations , as to suggest a systematic error of this kind .
Assuming that the error might amount to per cent. in the mean specific heat at C. with the thin bulbs , and that it would probably increase in proportion to the temperature and to the relative thermal capacity of the bulbs , it would amount to per cent. at C. , which would be more than sufficient to brin the results of DIETERICI into agreement with the most probable reduction of REGNAULT'S observations as indicated by my formula .
A similar uncertainty would apply with greater force to the experiments at higher temperatures where the thermal capacity of the quartz-glass bulbs amounted to four times that of the contained water .
The heat-loss in transference might have been in part eliminated from the results for water by using the same bulbs full and empty at each temperature , but even in this case the accuracy of the results for water would have been reduced to about a fifth with the thickest bulbs .
The large correction for the water equivalent of the bulbs , which could not easily be reduced , is a serious objection to DIETERICI'S method as compared with REGNAULT'S at the higher temperatures .
Below C. this source of error is unimportant as compared with evaporation losses incurred in transferring hot water when exposed to evaporation , as in LUDIN'S method .
On this account there would probably be little hesitation in preferring DIETERICI'S results to LUDIN'S between C. and C. , if it were not that , within the last year , Messrs. W. R. and W. E. BOIJSFIELD ( ' Phil. Trans. Roy .
Soc , 1911 , vol. , pp. 199-251 ) have succeeded in reproducing LUDIN'S results with remarkable fidelity by a method of electric heating with a vacuum-jacket calorimeter , which presents many ingenious and novel features .
Owing to the uncertainty in the reduction of DIETERICI'S results for the specific heat at C. , it might naturally be argued that his curve for the mean specific heat should be fitted to LUDIN'S at a higher temperature , such as C. or C. This could easily be done by raising all the points representing DIETERICI'S observations in fig. 2 by only per cent. , in which case they would nearly all agree with LUDIN'S curve to 1 in 1,000 except those below C. , where DIETERICI admits a larger possible error .
It would then appear that Messrs. BOUSFIELD , , and DIETERICI were in fair agreement in assigning a much higher value to the mean specific heat from C. to C. than that assigned by the continuous-electric method , and that the error probably lay with the latter .
It is usual to take the curves for the actual specific heat in making these comparisons , but these do not represent the observations themselves , except in the case of the continuous-electric method , and little can fairly be deduced from such a comparison ( though DIETERICI , LUDIN , BOUSFIELD , and many others have adopted this method ) , owing to the great uncertainty involved in deducing the from the mean specific heat .
There is no similar uncertainty in deducing the mean from the actual specific heat , so that the method adopted in fig. 2 is the more appropriate .
Since the evidence for the slow rate of increase of the specific heat between C. and C. , in the continuous-electric method , rested chiefly on half-a-dozen observations taken under conditions of exceptional difficulty , it appeared desirable to confirm them , if possible , over this range by an entirely different method , at least equal in accuracy .
Continuous-Mixture Method .
If two steady currents of fluid at different temperatures are passed through an arrangement of concentric tubes called a " " heat exchanger it is clear that , neglecting external heat-loss or gain , the loss of total heat by the hot current will be equal to the gain of total heat by the cold current .
By measuring the currents and the temperatures of inflow and outflow , we have all the data required for determining the ratio of the mean specific heats over the respective ranges .
This method does not appear to have been applied to any extent in accurate calorimetry , on account of the experimental difficulties involved in regulating and measuring the currents and the temperatures simultaneously to a sufficient order of accuracy .
As applied to the variation of the specific heat of a single fluid , the method permits of a most important simplification which does not appear to have been hitherto noticed .
In place of employin two separate currents , each of which must be measured and regulated to the limit of accuracy , the same current is passed twice through the heat exchanger , first as a hot current , and then , after suitable cooling , as a cold current , or vice versed .
If there is no leakage the ratio of the currents is always one of equality , and a comparatively rough determination of the absolute value of the current suffices for the application of small corrections .
The experimental problem is reduced to the regulation and measurement of the temperatures , which , taken by itself , is comparatively easy .
The method possesses the advantage , common to all continuous-flow methods , that a knowledge of the water equivalent of the calorimeter and of its variation with temperature is not required provided that the conditions are fairly steady .
There is no uncertainty of heat-loss in transference , or by evaporation , as with an open calorimeter .
It is also easy , by a suitable arrangement of the flow- tubes , to reduce the external heat-loss almost to a vanishing quantity without the employment of vacuum-jackets or elaborate precautions in lagging .
The method HEAT OF WATER , WITH EXPERIMENTS BY NEW METHOD .
requires considerable variation in detail to suit different ranges of temperature , and is in some ways less direct than the electric method , but it is peculiarly suitable as a means of independent verification .
eneral Arrangement of the Apparatus .
The continuous-mixture method was primarily designed for determining the variation of the total heat of water at temperatures above C. , where the continuous-electric method appeared to present greater difficulties .
But , as the Fig. 4 .
arrangements for this were not complete , it was decided to apply the method in the first instance as an independent test of the accuracy of formula ( 6 ) over the range C. to C. , where it differed most widely from those of LUDIN and BOUSFIELD .
The arrangement adopted for this purpose will be readily understood from the diagrammatic scheme in fig. 4 .
The current of air-free distilled water is supplied from a boiler in which the level is kept constant by means of a float-feed , from an auxiliary tank , maintained at a temperature of about C. From the boiler the current passes with a of about 15 feet to a reservoir surrounded by a steam-jacket , in which it is again raised to the boiling-point before entering the exchanger E. There is an air-trap at the top of the reservoir , and similarly at each of the highest points of the circulating system , to facilitate filling the apparatus completely with water before starting the flow .
Just before entering the exchanger , the current passes through thermometerpocket , the construction of which is shown on a larger scale in the section of the exchanger , fig. 5 .
In order to protect the thermometer-pocket itself from loss of heat , the current is made to circulate upwards through ajacket-tube surrounding the thermometer-pocket before circulating downwards past the thermometer .
Any small loss of heat which may occur before reaching the thermometer is immaterial .
The thermometer-pockets and other tubes throughout the exchanger are made of thick copper , with deep screw-threads carefully fitted to produce a spiral S circulation of the current .
After passing thermometer the current flows downwards through the central tube to hermometer Tpocket constructed shown iorder tvoid confusion .
TheIts temperature oirst 1eaving txchanger iaken boncentric twhere ioses hurrentthe bottom oxchanger ahence upwards through thermometer-pockets are all connected to the exchanger by thin tubes about 2 cm .
to 3 cm .
long and cm .
bore .
These connections are exaggerated in the diagram , fig. 4 , so as to permit the general course of the circulation to be more easily followed .
After leaving at a temperature between C. and C. , the current passes through a cooling-tube surrounded by an easily regulated jacket Fig. 5 .
of cooling water ; thence through a long tin spiral immersed in a large cooling tank to steady the temperature , before reaching a similar spiral in the regulator tank , where its temperature is reduced precisely to that required for the cold current .
From the regulator tank the current passes directly through the thermometer-pocket before entering the exchanger again as a cold current .
In order to prevent loss of heat from the exchanger to the regulator tank in which it is immersed , the current , at very HEAT OF WATER , WITH EXPERIMENTS BY A NEW METHOD .
nearly the same temperature as the regulator tank , first circulates downwards through the outer jacket and then upwards through the next inner tube , where it comes in contact with the hot current , emerging finally through the thermometerpocket , whence it passes to the collecting bottle , where the flow is measured from time to time .
The head being maintained constant , steady currents of suitable values are obtained by fitting exit tubes of various bores between and the collecting bottle .
From the collecting bottle the water is continuously returned to the heater on the floor above by means of a small rotary pump .
With the exception of the heat exchanger , the details require considerable modification for different ranges of temperature .
But the arrangement above described has been found to work very well for comparing the mean specific heat from C. to C. with the mean specific heat from C. to C. , and will sufficiently illustrate the general nature of the method .
With a flow of 10 .
per second the heat-exchallge amounts to about 300 calories per second , and the external heat-loss with the arrangement aboye described can be reduced to less than a tenth of a calorie per second , or about 1 in 3,000 of the total quantity measured .
Owing to the relatively small thermal capacity of the exchanger , and to the fact that the distribution of temperature is nearly independent of the flow , the conditions become steady to C. in a few minutes when the flow is changed .
The accuracy attainable depends chiefly on the limit of accuracy in reading the thermometers .
The Platinum Thermometers .
The thermometers employed were of my usual pattern in glass tubes , with leads partly of silver and partly of platinum , insulated by mica discs , spaced at intervals of 2 cm .
throughout the length of the tube by means of mica crosses .
Spacin the discs by means of mica crosses appears preferable to spacing the discs by long thin tubes of biscuit porcelain , as commonly practised by many makers , because the tubes are more hygroscopic than the mica .
They also make the compensation less sensitive by shielding the leads .
The thermometer coils were of pure standard wire , cm .
diameter , and the ends of the compensating were connected by fusing on a short piece of the same wire to eliminate any conduction effects which might exist .
Thick platinum leads extended for a distance of 7 cm .
from the coil , where they were fused to silver leads .
The object of this is partly to avoid possible contamination*of the fine wire with silver and partly to diminish conduction along the tube near the bulb .
The immersion of the thermometers in the apparatus was * From the first I have always adopted this method of uction in my own thermometers for accurate work at high temperatures .
It seems likely that many of the small variations of zero and difference-coefficient , found by careful observers , are due to contamination of the fine wire with gold or silver solder at its junction with the or silver leads .
HOLBORN and , in their recent papel ( loc. attribute of the small variations of their platinum thsrmometers to this causa VOL. CCXII.\mdash ; A 15 cm .
At the head of each thermometer the silver leads were soldered to flexible copper leads , 3 .
long , with amalgamated copper terminals fitting the mercury cups of the compensated box described in a previous paper CALLENDAR , .
This was essential to permit of rapid interchange of the thermometers in taking readings .
The ice- and steam-points were observed on several occasions .
The sulphur-point was also observed on one occasion , when the following readings were taken:\mdash ; Readings at the fixed points were taken and corrected for box-temperature and coil-errors to one figure beyond that given in the above table ( corresponding to but the last figure has been rejected as not being fully significant even at the fixed points , and as bein beyond the limit of accuracy at the sulphur-point .
The fundamental intervals of the thermometers were approximately 5 ohms each , or 500 cm .
of the bridge-wire .
Thermometers and had been made at the same time , by my assistant Mr. W. J. COLEBROOK , for use as a differential pair , and had been adjusted with special care so that their uncorrected readings agreed to 1 in 10,000 throughout the scale .
and were precisely similar thermometers made at different times , with resistances adjusted to about 1 in 1,000 .
The readings at the sulphur-point were taken in succession on the same day , when the barometer was falling slightly , in the order .
The corresponding temperatures , , on the gas-scale are calculated from the observed barometer readings by assuming the normal boiling-point to be C. , and the pressure variation to be C. per mm. , as found by CHREE .
The results show that all four thermometers agreed in their temperature scales to C. at the sulphur-point , which is nearly the limit of error of the readings .
An iron-tube apparatus was employed for boiling the sulphur and the thermometers were fitted with a single screen .
Experience has shown that this apparatus gives consistently a temperature nearly C. lower than the glass-tube apparatus with two screens , as originally described .
Allowing for this , the difference coefficients of all four thermometers would be within 1 in 1,500 of the standard value , , for pure platinum when the S.B.P. is taken as 44 53 C. In any case , this correction from to would not affect the scale of the thermometers by so much as C. at C. In my experience , platinum HEAT OF WATER , WITH EXPERIMENTS BY A NEW METHOD .
thermometers , when carefully constructed of suitable wire , always agree so closely with this scale that it is seldom worth while to make a special determination of the difference-coefficient by reference to the sulphur boiling-point .
It was , however , considered desirable to make the test in the present instance because the apparent variation of the specific heat of water depends in a great measure on the thermometric scale employed .
The thermometers were also compared differentially between C. and C. and were found to agree so closely that no differences in their scales could be detected .
Method of Reading the Thermometers .
The galvanometer employed was of the moving coil type , with a esistance of about 13 ohms .
Readings were generally taken by the constant-current method which I first introduced in 1891 .
In this method the platinum thermometer is connected on the compensator side of the box in series with the box coils , and is balanced against a compensated resistance , to , or greater than , the resistance of the thermometer at the highest point of the range it is desired to cover , in this case ohms .
With this arrangement the current through the thermometer is nearly the same at all temperatures , since the resistances on both sides of the bridge are kept nearly constant .
The current through the thermometer was approximately ampere , and was adjusted so that the deflection of the galvanometer was 1 mm. at 1 metre on reversal for a change of temperature of C. Readings were taken by setting the contact-point on the bridge-wire as exactly as possible with a lens to the nearest millimetre , and then observing the small deflection of the galvanometer on reversing the battery .
The definition was so perfect that it was possible to read to mm. of the scale , or C. , by the galvanometer deflection when the temperature was steady .
But this was the case only at the fixed points .
No attempt was made to read nearer than C. at other temperatures , but the provision of ample sensitiveness greatly facilitated quickness and certainty of reading .
The constant-current arrangement of the measuring apparatus possessed special advantages for the continuous-mixture method , because it was necessary read four * These compensated resistances are very useful in accurate resistance measurements or platinum thermometry .
They are best constructed of platinum-silver , wound on mica and annealed in situ at a dull red heat , compensated by a small resistance of pure platinum , having an equal temperature increment , and connected in the opposite arm of ths bridge ( CALLENDAR , 'Brit .
Patent , ' No. 14,509 , 1887 ) .
The two coils may conveniently be enclosed in a glass tube with compensated leads like a platinum thermometer , and the point of zero temperature-coefficient may easily be adiusted as near to C. as desired .
In this case the whole change of resistance of the combination between C. and C. is less than 1 in 100,000 , and the change between C. and C. is only 1 in 2,000,000 .
The best specin1ens of manganin generally have a temperature-coefficient of 2 or 3 in 100,000 C. at C. , which is more than 100 times as great ; but even if the point of zero-coefficient manganin couId be djnsted to C. , the changes in the neighbourhood of C. would be more than 10 times as grcat as with the platinum-silver coil compensated with platinum .
For a further discussion of this point see NDAR , ' Phil. Trans , vol. 199 , p. 89 , 1902 .
PROF .
H. L. CALLENDAR ON THE yARIATION OF THE SPECIFIC thermometers at different temperatures in rapid succession .
It was important that the sensitiveness of the galvanometer should be the same for each , and it was desirable that the rise of temperature produced by the measuring current should be nearly the same for each thermometer .
This last condition is approximately secured by keeping the current constant .
CHAPPUIS and HARKER ( ' Phil. Trans , 1900 , p. 62 ; CALLENDAR , .
cit. , p. 93 ) proposed to do this by keeping the watts constant , adjusting the cur \amp ; nacute ; t to suit the value of R. But the emissivity of the wire increases somewhat more rapidly than , so that the rise of temperature due to at different points of the scale is nearly proportional to .
The rise of temperature produced by a current of ampere at C. was measured and found to be C. The rise at C. was found to be C. Assuming that the variation between these limits was regular , it was evident that it could not produce a systematic error of the temperature scale greater than C. between the limits C. and C. In measuring the mean specific heat over a range of C. , limit of accuracy of C. in the thermometric readings appeared to be ample , because this would amount to only 1 in 30,000 of the heat measured , and it was hardly to be expected that the external heat-loss could be determined with a much higher order of accuracy than 1 in 10,000 .
Theory of the Continuous-Mixture Method .
If X is the external heat-loss in calories per second , and the water current in grammes per second , the equation connecting the mean specific heats over the ranges to and to for a single value of the flow is evidently . . . . .
( 7 ) If the heat-loss X could be neglected by sufficiently increasing the flow , this equation would give the required ratio of the specific heats directly , bein simply the inverse ratio of the temperature ranges .
In any case , if X is small and large ( say 10 to 20 , this would give a good first approximation , better than 1 in 1,000 if X is less than 1 in 1,000 of the whole heat exchange .
Assuming that the temperature distribution in the exchanger , and consequently the heat-loss X , does not vary appreciably when the flow is changed within reasonable limits , a second approximation could easily be secured by employing the first approximation to evaluate the heat-loss for a small flow , say 1 , and employing the value so obtained for the large flow ; or the heat-loss X might be directly eliminated by subtracting one equation from the other if the temperature ranges were so nearly the same that the values of the mean specific heats could be assumed to be the same for the small flow without sensible error .
This method of reduction would undoubtedly give good results if the losses were smau .
In practice , however , it is impossible to secure exact similarity in the temperature distribution for flows varying in the ratio of 10 to 1 , and it is , HEAT OF WATER , WITH EXPERIMENTS BY A NEW METHOD .
therefore , preferable to adopt a method of reduction depending on some assumed variation of the total heat .
This appears at first sight a less direct method , but is peculiarly appropriate when the primary object of the experiment is to verify formulae already obtained by different methods .
Variation of the I'otal Heat .
The variation of the total heat is not so familiar as the variation of the specific heat , but since the change of total heat between given limits is the quantity actually measured in a calorimetric experiment , the total heat is generally the most useful quantity to tabulate for experimental purposes .
The numerical value of the total heat from C. to C. in terms of a unit at C. differs but little from over the range C. to C. It is , therefore , convenient to write , . . . . .
( 8 ) where is the small excess of over at any temperature , which may appropriately be called " " the variation of the total heat The value of given by LUDIN'S formula ( 5 ) is , ( 9 ) whence the value at C. is The corresponding formula for deduced from my formula ( 6 ) representing the results of the continuous-electric method is , ( 10 ) whence the value at C. is , differing from LUDIN'S formula by nearly per cent. These two formulae are represented by the curves in fig. 6 .
In order to save space , Fig. 6 .
LUDIN'S curve from C. to C. is represented in three pieces , the curve being shifted downwards through when it reaches the upper limit of the diagram .
ROWLAND'S observations on the mechanical equivalent from C. to C. , if plotted on the same scale , would agree to throughout his range with formula ( 10 ) , the full curve .
His experiments do not , as generally stated , afford any conclusive evidence of a minimum at C. in the specific heat .
ROWLAND himself considered that , owing to the increasing magnitude and uncertainty of the radiation correction beyond C. , " " there might be a small error in the direction of making the mechanical equivalent too great at that point and the specific heat might keep on decreasing to even The discrepancy from LUDIN'S curve in this region is less than 1 in 1,000 of the total heat , and is within the limits of error of LUDIN'S experiments .
Formula ( 6 ) gives a minimum at C. LUDIN'S formula gives a minimum at C. , which cannot be reconciled with ROWLAND'S observations .
The point at which the specific heat in absolute units can be most accurately inferred from ROWLAND'S experiments is C. , at the middle of his temperature range .
The value at this point is probably within 1 in 2,000 of joules per .
C. on the scale defined by formula ( 1 ) .
ROWLAND himself gave the value on the air scale ; reduction of ROWLAND'S thermometers gave on the hydrogen scale .
The results of REYNOLDS and MOORBY , when corrected to cover the range C. to , give joules for the mean calorie .
Combining these two results we find , for the ratio of the mean calorie to the calorie at C. My formula ( 10 ) would make the ratio , which is fairly good agreement .
But LUDIN'S formula gives the ratio , showing a discrepancy of per cent. , which would appear to be beyond the possible limits of error of REYNOLDS and MOORBY'S experiments , since the extreme variations of their results did not exceed per cent. , and the mean would probably be correct to 1 in 1,000 .
The evidence to be deduced from the observations of W. R. BOUSFIELD and W. E. BOUSFIELD ( loc. cit. ) with regard to the variation of the total heat rests on their " " point-to-point\ldquo ; experiments over the following ranges:\mdash ; Range of temperature C. C. C. C. C. C. Mean value of J. deduced . . . .
LUDIN . . . . .
The values of the variation at the points C. , C. , C. , C. , and C. are deduced from Messrs. BOUSFIELD'S observations by taking the-ir value for the unit at C. and adding up their values of the total heat for the separate ranges .
Up to C. they agree very closely with LUDIN'S curve , as indicated by the crosses in fig. 6 .
Their values at C. and C. are somewhat HEAT OF WATER , WITH EXPERIMENTS BY A NEW METHOD .
lower than LUDIN'S , but the discrepancy here is only from C. to C. They considered that the measurements could not be carried beyond this point with an open calorimeter owing to the rapidly increasing uncertainty of the heat-loss due to evaporation , " " obturator\ldquo ; heating , \amp ; c. The general agreement with LUDIN'S curve is so striking that it would seem at first sight as though it could hardly be due to chance .
Their methods had no points in common , except that they both used mercury thermometers and open calorimeters .
I\amp ; duction of Observations by the Total-Heat fethod .
The curves of variation of total-heat above described and explained , when plotted on a scale of 1 mm. to C. for , afforded the most accurate and expeditious method of reducing the observations by the continuous-mixture method , since it was easy to take exact account in this way of small differences of range with different flows .
If are the total heats from C. corresponding to the observed temperatures , equation ( 7 ) , when expressed in terms of the total heat , becomes . . . . . .
( 11 ) Substituting , and writing for the sum , and for the sum , this equation may be written . . . . . .
( 12 ) If the formula assumed for is correct , the values of the heat-loss X deduced different flows should be nearly the same for the same temperature range , or should vary in relation to the mean excess-temperature of the calorimeter ) its surroundings .
If the formula employed for is incorrect , the calculated values of the heat-loss X will vary with the flow , and will bear no consistent relation to the external conditions .
As there was some delay in procuring the requisite sizes of copper tube for the high-pressure apparatus designed for work at temperatures up to C. , the apparatus was in the first instance constructed of ordinary brass tubing with solder joints for work below C. The soft-solder joints would not stand high temperatures , but had the advantage that the apparatus could be takell to pieces readily and remade in different forms , to test the effect of variations in the external conditions on the heat-loss observed .
Thus in order to test for possible due to conduction of heat along the thin tubes , 2 to 3 cm .
long , 5 mm. bore , and thick , connecting the thermometer-pockets to the exchanger , the connecting tubes were replaced by short thick brass cylinders 1 cm .
long by 2 cm .
in diameter .
This produced an appreciable effect on the thermometer readings only with the smallest flow of .
It was inferred that the conduction error with the thin tubes actually employed would be inappreciable even with the smallest flow .
A possible source of error to be carefully avoided in the continuous-mixture method is leakage of fluid at any point of the circuit between the first and last thermometers .
Precautions against this were taken in designing the apparatus by making the cooling coils and other inaccessible parts of the apparatus of continuous lenghs of solid-drawn tubing , and arranging all joints and connections as far as possible so as to be easily accessible and at all times open to inspection .
In addition to this , a leakage test was regularly applied each day before starting observations , when the apparatus was at a nearly uniform temperature .
The boiler was disconnected , and a glass tube of small bore was attached to the inflow at the top , to serve as an index .
Any possible leaks were thus subjected to the full head of 15 feet .
A special test was also made for leakage between the hot and cold tubes in the exchanger .
A small continuous leak due to strain , amounting to about 1 .
per hour , was detected on one occasion , and immediately rectified .
But , as a rule , the movements of the sensitive index were negligibly small , being due merely to slight changes of temperature , and more often positive than negative .
The apparatus was also tested with and without the jacket-tube , figs. 4 and 5 , between the exchanger and the regulator tank in which it was immersed .
The loss from the thermometer-pockets , when unjacketed , was found to be from to calories per second , according to the conditions of the experiment , and to be nearly independent of the flow , since the mean excess-temperature of the thermometerpockets above the regulator tank , namely , varied but slightly with the flow .
The loss from the inner tubes of the exchanger itself , when unjacketed , was nearly 2 calories per second , but varied by nearly 20 per cent. of itself when the flow was changed in .
the ratio of 1 to 10 .
This could easily be foreseen , because the cold current was raised to a higher temperature when the flow was small .
The actual distribution of temperature in the outer tube of the exchanger was observed by means of thermocouples , and found to correspond with theory and with the observed changes in the heat-loss with flow .
The jacket-tube was designed to intercept this variable heat-loss and reduce it to about a tenth for the smallest flow .
The mean rise of temperature of the jacket-tube for the smallest flow was found to be of the order of C. The mean difference of temperature between the jacket-tube and the regulator was always measured and allowed for in estimating the mean excess-temperature , but it was generally less than C. for the largest flow , corresponding to a heat-loss of about 1 in 10,000 , which was about the limit of agreement of the temperature measurements .
It was ultimately decided to do without jackets for the thermometerpockets at temperatures below C. , because the heat-loss from the thermometerpockets at these low temperatures was fairly small and nearly independent of the flow .
The omission of these jackets considerably simplified the construction and connection of the apparatus .
The main jacket for the exchanger was employed in nearly all the experiments , because it made the heat-loss so nearly independent of the flow , and made the reduction of the observations comparatively simple and certain .
HEAT OF WATER , WITH EXPERIMENTS BY A NEW METHOD .
For regulating the temperature of the cold current , an electric temperature regulator was employed of the same type as that used in the continuous-electric method , which had been specially designed to prevent hunting .
With a .
lamp , and suitable adjustment of the cooling water , currents up to 20 .
could be kept steady to a few thousandths of C. for considerable periods , over a range from C. to C. But this was not such an easy matter as might be imagined , owing to variations of temperature of the cooling water .
During the months of May , June , and July , 1911 , upwards of 150 complete sets of readings were taken , each for a single flow .
The upper limit of temperature was always nearly C. , but the lower limit was varied from C. to C. The flow was varied from 1 .
to 20 .
It is a great advantage of the method that it permits the flow to be so widely varied without introducing any serious experimental difficulties , but the largest flows were somewhat difficult to regulate , and there was no material gain in accuracy beyond about 10 , at which point the limit of accuracy of the thermometric readin gs corresponded with the limit of reading the total heat curves .
Many variations were tried in the disposition of the apparatus , with corresponding variations in the heat-loss , which , even without jackets or lagging , seldom exceeded 1 per cent. of the heat-exchange for the largest flow .
The results of all these experiments were so nearly similar that it will suffice to give one example of the detailed readin gs .
The following readings were taken with the jacket-tube surroundin the exchanger , but without or jackets on the thermometer-pockets :\mdash ; OBSERVATIONS of July 3 , 1911 .
Thermometer from curve The above table includes observations for five different values of the flow .
Each flow was maintained steady for about half-an-hour , during which time two measurements of the flow were taken , and four sets of readings of the thermometers .
The flows generally agreed to 1 in 500 , and the thermometers seldom varied more than C. in half-an-hour .
In the above example the readings for the maximum flow were continued for over an hour .
Such changes in the thermometer readings as occurred during each flow were due partly to change in the barometer and pal.tly to change in the temperature of the cooling water , which also made it necessary to reset the regulator between each flow , as indicated by the readings of the lneter t VOL. CCXII.\mdash ; A. PROF .
H. L. CALLENDAR ON THE VARIATION OF THE SPECIFIC lt was found on reducing the above observations that insufficient time had been allowed for the conditions to become steady after changing the flow from to .
A change of C. occurred in after the first pair of readings at .
All the readin gs have been included in the mean , but , if the first pair had been rejected , the heat-loss would have been for this flow .
The signs attached to the temperatures are those taken in finding the sum .
The values of were taken from the curve given in fig. 6 , representing my formula ( 10 ) .
The values of the heat-loss X deduced are nearly the same for all the flows .
It should be observed that the heat-loss , , is less than per cent. of the heat-exchange , 325 cal .
, for the large flow , and that the flow could easily be varied in the ratio of 1 to 8 .
In the continuous-electric method with a vacuum-jacket , the heat-loss at C. amounted to 4 per cent. of the maximum watts , or 10 per cent. of the difference between the flows , on which the result depends .
In REYNOLDS and MOORBY'S experiments the heat-loss amounted to 5 or 10 per cent. ( with or without lagging ) of the difference of the loads in the heavy and light trials .
In neither case could the flow be varied satisfactorily in a ratio greater than about 1 to 2 .
The continuous-mixture method is undoubtedly preferable to either in this respect , since it permits a wider range of variation of the flow , and a greater reduction in the heat-loss .
The agreement of the values of the heat-loss deduced from the different flows by means of formula ( 10 ) is closer than might have been expected , because 1 in the last figure of the heat-loss corresponds to C. , or 1 in 30,000 of the heatexchange for the large flow .
It may be said that formula ( 10 ) is verified to at least 1 in 5,000 for the ratio of the mean specific heat from C. to C. , where it differs most widely from LUDIN'S , to the mean specific heat from C. to C. Formula ( 10 ) gives for the ratio .
LUDIN'S formula gives , differing by per cent. If LUDIN'S formula had been employed for the reduction , the heatloss , instead of bein nearly the same for the different flows , would have appeared to vary from for the largest flow to 179 for the smallest flow .
The heat-loss should , as a matter of fact , have been slightly less for the large flows than for the small , because the rise of temperature of the jacket with the smallest flow was rather more than sufficient to compensate for the fall of mean temperature of the thermometer-pockets .
Seeing that the results of the continuous-electric method have now been so closely verified by the continuous-mixture method , which is independent of electrical energy measurements , it would appear to follow that the discrepancy of 1 per cent. at C. between these methods and those of Messrs. BOUSFIELD and LUDIN is to be attributed mainly to fundamental differences in the thermometric and calorimetric methods employed .
In my continuous-flow methods the troublesome and uncertain corrections of mercurial thermometry at temperatures between C. and C. have been avoided , and a higher order of accuracy in the temperature measurements has been secured by the direct employment of platinum thermometers .
Errors due to lag , or HEAT OF WATER , WITH EXPERIMENTS BY A NEW METHOD .
to uncertainty of the water-equivalent of the calorimeter and its variation with temperature , have been practically eliminated by keeping all the conditions steady , so that the observations could be pushed to the limit of accuracy of temperature measurement .
Errors due to dissolved air and to evaporation , which are quite appreciable at C. , and begin to be serious at C. , with an open calorimeter , have been minimised by keeping the water free from air up to the moment of its passage into the calorimeter , where it is completely enclosed and protected from evaporation .
Without presuming to criticise in detail the work of Messrs. BOUSFIELD , I maintain that these difficulties have not been adequately met in their investigations , and would inevitably give rise to constant errors , which could not be detected without fundamental variations in the conditions of experiment .
While their method may undoubtedly be suitable for the purpose for which it was originally devised , namely , that of comparing the specific heats of similar solutions over the same range of temperature , I cannot admit that it affords any promise of exceptional accuracy in the determination of the variation of the specific heat over different ranges of temperature , which is a much more difficult problem .
According to my ow1l work and that of Dr. BARNES , the whole variation of the specific heat of water between C. and C. is less than one-half of 1 per cent. , and demands the most accurate methods of investigation .
The occurrence of so high a maximum as LUDIN'S experiments show in the neighbourhood of C. is theoretically inadmissible , and caImot be reconciled with the work of REYNOLDS and MOORBY , or with the corrected results of REGNAULT , which are satisfactorily represented by my formula .
Variation of Heat with The continuous-electric and continuous-mixture methods both compare the variations of the total heat of the fluid , , under the condition of practically constant pressure , or give ratios of the values of the mean specific heat at constant pressure over different ranges of temperature .
Below C. the pressure is always atmospheric , and the variation of the specific heat with pressure does not exceed 1 in 10,000 per atmosphere .
At higher pressures and temperatures it would be necessary to take account of the variation of the specific heat with pressure in comparin observations taken at different pressures , since the pressure in the apparatus must always exceed the saturation pressure at the highest temperature observed .
This reduction can be effected with sufficient ximation by means of the thermodynamical relation , The variation amounts to nearly 2 in 10,000 per atmosphere at C. The quantity of heat supplied to water in a boiler , maintained at constant pressure and temperature , per unit mass of water pumped into it at the same pressure but at a lower PROF .
H. L. CALLENDAR ON THE yARIATION OF THE SPECIFIC temperature , is the change of total heat at constant pressure , and is a quantity of the same kind as that measured in the continuous-mixture method .
The work , which is generally included in tabulated values of the total heat , is supplied as work by the feed-pump , and not as heat in the boiler .
It would amount to only 1 in 500 of the total heat at , and might be neglected except in the most accurate calorimetric work .
The quantity measured in REGNAULT'S method was the change of total heat from boiler pressure and temperature down to the final temperature of the calorimeter at atmospheric pressure , which may be denoted by the expression , where the suffix 1 denotes atmospheric pressure at the final temperature of the calorimeter .
The differential of this is the rate of change of the total heat of water under saturation pressure , and exceeds the specific heat at a constant pressure equal to that of saturation by the expression , which amounts to nearly per cent. at 20 C. The specific heat at constant pressure exceeds the so-called " " saturation\ldquo ; specific heat at the same pressure and temperature by the expression , which amounts to per cent. at C. The values of the total heat and the specific heat from C. to C. given in the following tables , in so far as they represent a reduction of REGNAULT'S observations , must be taken as representing the total heat of water and its rate of variation under saturation pressure , and not under constant pressure ; but the uncertainty of the reduction probably exceeds the difference in question .
The quantity measured by DIETERICI was the change of intrinsic energy of water from saturation pressure and temperature in the heater to saturation pressure at C. , which may be denoted by .
Neglecting the small change of intrinsic due to 1 atmosphere at C. , the quantity measured by DIETERICI is less than that measured by REGNAULT by the expression , which amounts to 043 calorie at C. , or per cent. on the mean specific heat .
His tabulated values of the mean specific heat at 20 C. , when reduced to a unit at C. , are already per cent. higher than my reduction of REGNAULT'S observations , so that the discrepancy is increased to 1 per cent. at this point if allowance is made for the difference in the quantities measured .
The actual specific heat at tabulated by DIETERICI is the rate of increase of the intrinsic energy under saturation pressure , which is less than that of the total heat under the same conditions by the expression , which amounts to nearly 1 per cent. at C. The specific heat tabulated by DIETERICI already exceeds that deduced from REGNAULT by 2 per cent. at C. , so that the discrepancy is increased to 3 per cent. in the value of the actual specific heat at this point .
The discrepancy is of the same order as the whole variation of the specific heat , and is not unimportant from a theoretical standpoint .
Accurate at these temperatures are very difficult , but it is clear that further experiments are desirable , if any theory of the variation of- specific heat is to be framed or tested , HEAT OF WATER , WITH EXPERIMENTS BY A NEW METHOD .
Explanation of the Tables .
The most useful table in practice is that of the variation of the total heat , which also permits the mean specific heat between any limits to be readily calcnlated .
The corresponding curve , representing the variation between C. and C. , has been given in fig. 6 , but the scale of fig. 6 does not permit the values to be read with sufficient accuracy from the curve as reproduced .
It should be observed that it is important to tabulate the variation in terms of the specific heat at C. taken as unity , and not in terms of the specific heat at C. , which is so often taken as the standard , because in the latter case the values of from C. to C. would all be negative , which would be inconvenient in using the table .
If , on the other hand , the minimum value of , or the value at C. , were taken as the standard , the values of would be inconveniently large .
The table gives the values of for each degree , and a column of differences is added to facilitate interpolation if desired , but the differences are so small for the greater part of the range C. to C. that this is seldom required .
Above C. the differences are larger , but the values are here so uncertain that it could seldom be worth while to interpolate .
The method of using the table is fairly obvious , but the following examples may make it clearer .
To find the total heat from C. to any point ; add to the exact value of expressed to C. , the corresponding value of for the nearest whole taken from the table , interpolating for fractions of a degree if great accuracy is required in a problem depending on small differences .
Unless is known to C. , interpolation is unnecessary .
To find the change of total heat between and ; find from the table the corresponding values of , namely , and , and add the difference to the difference , with due regard to sign .
To find the mean specific heat from C. to ; divide the corresponding value of by and add unity .
To find the mean specific heat between and ; find the ence d , divide by the difference , and add unity .
Thus , if the given values are C. , C. , we find , whence .
The result will be correct to 1 in 10,000 , if is not less than C. If the range is less than C. , the specific heat at the mean poin of the range , taken from the table of specific heat at , is sufficiently close approximation in most cases .
The values of the entropy of water reckoned from C. are sometimes required , and are generally given in steam-tables .
Assuming that C. is C. from the absolute zero , the formula for the entropy obtained by integrating from C. to formula ( 6 ) for the specific heat divided by is as follows : HEAT OF WATER , WITH EXPERIMENTS BY A NEW METHOD .
PROF .
CALLENDAR ON yARIATION OF SPECIFIC HEAT OF WATER , ETC .
If the specific heat were constant and equal to unity the value of would be The values of the specific heat , total heat , and entropy between C. and C. are for a constant pressure of 1 atmosphere .
The values above 10 C. are for water under saturation pressure .
The difference between saturation pressure and 1 atmosphere would not affect the values by 1 in 10,000 below C. In conclusion , my best thanks are due to Mr. W. J. COLEBROOK , Superintendent of the Physics Workshop of the Imperial College , and to other members of the staff under his direction , for the great care and skill displayed in making and adjusting the platinum thermometers and other apparatus required for this investigation .

