591 On the Observed Variations in the Temperature Coefficients of a Precision Balance .
By J. J. Manley , Hon. M.A. Oxon .
, Daubeny Curator , Magdalen College , Oxford .
( Communicated by Prof. E. B. Elliott , F.R.S. Received March 8 , \#151 ; Read April 25 , 1912 .
) Introductory.\#151 ; In a former paper* communicated to the Royal Society , the apparent abnormal behaviour of certain precision balances was described and discussed .
One of the most important points therein considered was that dealing with the displacement of the resting-point , R.P. , of a beam , consequent upon a uniform rise or fall in the temperature of the instrument .
The magnitude of the displacement of the resting-point , R.P. , for a change of 1 ' C. was termed the temperature coefficient of a balance .
The variations in the R.Ps .
of the balances tested were observed within a somewhat limited range of temperature only .
Within those limits they were found to be regular and expressible by an equation of the form R t \#151 ; Rs { 1 i \lt ; *t i ftt2 } , in which Rs and R* are the R.Ps .
at some standard temperature and t ' respectively , and \#171 ; and ft experimentally determined constants for a given balance and particular load .
In this present communication the author would direct attention to the hitherto unrecorded changes which may occur in the temperature coefficients of delicate balances , and offer some further remarks upon the subject in general .
, Description of Balance.\#151 ; The balance used during this investigation was obtained new , direct from the makers , Messrs. A. Gallenkamp and Co. , about three years ago .
The beam is cantilever in type and enclosed by an inner protecting case , as recommended in my former paper .
As the result of prolonged testing , I conclude that the balance fully merits the distinction which the makers claim for it , namely , a practically constant sensibility for all loads varying from 0 to 200 grm. , this latter being the maximum load for which the balance was built .
The sensibility of the balance was , with advantage , set somewhat high , and equal to 45 divisions per 1 mgrm .
; its value was kept unchanged throughout .
The equality of the temperature of the beam could at any time be tested by means of a differential bolometer * 'Phil .
Trans.\gt ; A , vol. 210 , p. 387 .
592 Mr. J. J. Manley .
Observed Variations in the [ Mar. 8 , placed within the beam case , and capable of indicating so small a difference as 1/ 10,000 ' C. As the bolometer was fully described in my former paper an account of it here is unnecessary .
Experimental Work.\#151 ; The temperature coefficient was , in the first instance , calculated from data obtained during March , 1909 , the range in the temperature during the observations being from 10 ' to 16 ' C. The accompanying graphs ( fig. 1 ) illustrate the variations which at that time occurred in the E.P. as the temperature , or load , or both were altered .
O- ( i m.fmrt or t Cantilever Beam March 1909 .
o-4 .
x 0-3 * Cf 02 \#171 ; v^y *0 ' ! .
0 nr ir n ' it H9 ip i6'C .
Fig. 1 .
As the variations in the R.P. with changes in the temperature and load were found within the above-named limits of temperature to be both definite and regular , the conclusion was drawn that for a small additional range , the application of the method of extrapolation might be resorted to with a sufficient degree of safety .
This course was therefore adopted for reducing to a standard temperature a series of weighings , some of which had been 1912 .
] Temperature Coefficients of a Precision Balance .
593 carried out when the temperature of the balance room approached 18 ' C. It was then seen that , whilst the weighings at temperatures up to and including those at 16 ' C. were , when reduced , very concordant , those made when the temperature approximated 18 ' C. gave , when corrected and compared with the values obtained at the lower temperatures , inharmonious and variable values .
With the object of elucidating the cause of these discordant values , each pan of the balance was loaded with 130 grm. , this load being practically equal to that of the substances which were being weighed when the discrepancies were detected .
Taking the precautions shown to be necessary and described in my former paper , a number of observations of the R.P. were made at temperatures ranging from 16 ' to 21*6 ' C. , during August 1 , 3 , and 4 , 1910 .
The different values then found for the E.P. are plotted against the corresponding temperatures and shown graphically in fig. 2 .
H - 0-3 .
Cantilever Beam Loac/ =/ 30 grms. 5=45 0 2 . .
4 .
o\#177 ; .
/ - 4 , 1910 .
0 / 6 ' it / 8 ' 19 ' 20* 21 ' 22'C Fig. 2 .
Here it may be seen that the regularity which exists in the temperature coefficient curve up to 16*7 ' C. , or thereabout , suddenly disappears ; between that temperature and 19*0 ' C. the beam is , with reference to its temperature coefficient , in a state of unstable equilibrium .
We have here a well-marked example of what I venture to term a critical temperature range for this particular balance and load .
Referring again to the curve ( fig. 2 ) we observe that for temperatures above 19*0 ' C. the coefficient k again becomes well defined and regular ; but its former positive value has now given place to another and negative value .
594 Mr. J. J. Manley .
Observed Variations the [ Mar. 8 , Therefore , if an object is weighed when the temperature is below 16 ' C. , its weight will appear to increase as the temperature of the balance approaches the limit , 16 ' C. Similarly , there will be an apparent decrease in the mass of the same body when it is weighed after the temperature has passed beyond the upper limit , 19-0 ' C. All weighings taken when the temperature falls within those two limits will yield more or less uncertain values .
Having completed the observations just described , the balance was rested for two months , and then a new series of determinations of the R.P. was commenced .
These were conducted at intervals during October 4 to 8 inclusive , the temperature of the balance varying within the limits 14-7 ' and 206 ' C. The results obtained upon this occasion are represented in graph No. I* ( fig. 3 ) .
The beam was now given the more prolonged rest of 23 weeks , after Cantilever Beam Load=130 grms. S -4 Fig. 3 .
which a third series of observations was undertaken .
These were commenced on March 22 , and concluded on March 24,1911 .
During that time the temperature of the balance varied within the limits 12'9 ' and 19 9 ' C * The portion a ... / 3 was obtained by extrapolating .
1912 .
] Temperature Coefficients of a Precision Balance .
595 Graph No. 2 ( fig. 3 ) was plotted from the data secured during this last series of experiments .
Although the characteristics of the two curves are similar , it will be observed that the slope of No. II is less marked than that of No. I. Comparing these two curves with those given in fig. 1 , we also observe that the change from a positive to a negative value is accompanied by a decided simplification in the nature of the curve ; and that , wdiereas formerly the type of the temperature coefficient k was represented by the equation k = 1 +at\#151 ; fit2 , it has now , within the experimental limits , become expressible as a straight line equation , namely , k = 1 \#151 ; oij .
Since the completion of the third and last series of determinations from which the present value of k was calculated , a very large number of weighings of different objects have been carried out with this balance ; the temperature has been allowed to vary quite freely ; the maximum temperature was as high as 23*4 ' C. in August , and the minimum as low as 10*4 ' C. in October , 1911 .
During the past ten months the temperature coefficient k appears to have remained strictly constant ; the values obtained from numerous weighings of the several bodies at different temperatures having shown , when reduced to a standard temperature of 16 ' C. , a remarkable degree of concordance .
With the balance adjusted to so high a degree of sensibility as that already named , it is believed that any further change in k could scarcely have escaped detection .
Discussion.\#151 ; On a previous occasion* an attempt was made to explain why certain precision balances exhibited a behaviour which was termed anomalous , and the opinion was expressed that the observed and correlated variations in the temperature and E.P. of a beam were primarily due to very small inequalities in the movements of the several groups of screws , together with their associated knife-edge blocks .
The additional evidence secured during this and another research not yet concluded , tends to confirm and strengthen this opinion .
The probability of the correctness of such a view will , it is believed , be apparent from the following theoretical considerations .
If , when the temperature of the beam is varied , the nature and sum of all the resultant movements of the screws forming one group are precisely equal both with regard to the nature and sum of all the movements of the corresponding screws of an opposed similar group , there will result a small but perfectly symmetrical movement .
Such a movement may either lengthen or shorten the two arms of the beam , but will leave the ratio of their lengths unaffected .
Therefore , under these conditions , provided the masses of the disturbed parts are strictly equal , the E.P. of the beam will experience no change and the temperature coefficient k = 0 .
* ' Phil. Trans. , ' A , vol. 210 , p. 398 .
VOL. LXXXVI.\#151 ; A. 596 Mr. J. J. Manley .
Observed Variations in the [ Mar. 8 , If , however , the total movement of two opposed groups of screws is unsymmetrical , the final result will be a more or less distinct differential displacement of the one group with reference to the other , and the greater the change in the temperature , the larger will be the differential displacement .
Now any perceptible movement of this nature will at once affect the ratio of the two arms and thus cause a shifting of the E.P. This shifting , when it proceeds with uniformity as the temperature increases or decreases , may be determined and its value per 1 ' C. calculated .
A definite expression for the temperature coefficient k is thus obtained ; k may be either positive or negative according to the direction in which the E.P. moves as the temperature increases or decreases .
If this mode of accounting for the presence or absence of the temperature coefficient of a balance be rejected then it would appear that there still remain two alternatives .
1 .
It may be suggested that the variations in the E.P. are caused by slight differences in the flexure of the two arms as the temperature varies .
In my former paper it was shown that this view is incorrect , for in no investigated case was there found any measurable difference in the flexure of the two arms when the temperature or load , or both , are varied .
Further , the balance used during this present research maintained an unchanged sensibility for all loads ; this would have been an impossibility had the beam suffered either equal or differential flexures .
2 .
The remaining alternative assumes the possession by the beam of distinctly different coefficients of expansion for the two arms .
It is , however , well known that in the case of first-class precision balances , very great care is exercised both in preparing and selecting only such alloys , the physical characteristics of which suggest a highly homogeneous whole ; and , therefore , any conclusion based upon the assumption of the presence of irregularities in the coefficient of expansion of the beam would appear to be not only unwarranted but also untenable .
If these views be accepted , we meet with no insuperable difficulty in accounting for the observed transformation of a positive temperature coefficient into one which possesses a negative and distinctly different value .
We proceed to consider this point .
The transient nature of the instability observed in the E.P. of the cantilever beam as the balance was made to pass through the critical temperature range , may to some appear not a little remarkable .
But in the authors opinion this behaviour of the beam should cause no surprise .
We are here dealing with a balance which had been in use for a comparatively short period ; and , therefore , the conditions as to load and temperature had 1912 .
] Temperature Coefficients of a Precision Balance .
597 not been sufficiently varied to enable the knife-edges and their kindred parts to take up truly normal positions .
For the purpose of removing some slight abnormal strains it was imperative that the beam should be subjected to the combined influences of a certain minimum load and a range of temperature approximately equal to 3 ' C. above 16*7 ' C. As soon as those conditions were fulfilled , it appears that some one or more of the parts secured to the beam proper moved into other and slightly different positions ; these new positions they still retain .
Such slight readjustments in the relative positions of the several screw groups might readily be accompanied by a transference of a major strain from one group to another , and thus , in a perfectly natural manner , bring about the changes observed both in the sign and value of k as cited in the experimental section of this paper and shown graphically in fig. 2 .
Some may dissent from the views and conclusions expressed above , but few , if any , will deny the importance of the experimental results which have been obtained during this present research .
We therefore pass from points of theoretical interest to a consideration of the manner in which , according to our view , the beam must be " aged/ ' On the " Ageing " of a Beam.\#151 ; The marked changes noted in both the character and value of the temperature coefficient of the cantilever beam almost inevitably lead us to conclude that , before a new , or even comparatively new , precision balance can be safely used for securing trustworthy values of the highest order of accuracy , the instrument must be suitably " aged .
" In making this suggestion nothing novel or unusual is proposed .
Every physicist is familiar with the fact that many precision instruments* ( such as standard resistance coils and thermometers ) are incapable of yielding really reliable values before they have been subjected to an " ageing , r process of some kind . .
It is fortunate that in the case of a balance beam , the ageing can be so* conveniently effected .
If the single instance already described can be accepted as a safe criterion , it will appear that the ageing of the beam maybe brought about as follows:\#151 ; The balance should be first placed in a room having a somewhat lower temperature than that likely to be encountered during the subsequent research ; the pans must then be suitably loaded and the beam released .
( Possibly the load should be the maximum for which the instrument is* built ; but experiment alone can enable us to determine this .
) After a time the temperature of the room should be gradually increased until it has passed beyond that which is known to be the upper temperature limit .
If the R.P. be determined at intervals during the slowly increasing 598 Mr. J. J. Manley .
Observed Variations in the [ Mar. 8 , temperature , we at the same time obtain the necessary data for calculating temperature coefficient k of the beam .
By repeating both the process and the observations , we may acquire all the evidence we need for a critical examination of the behaviour of the beam during the whole treatment , and thus be able to determine whether a further ageing of the beam is desirable .
It is quite possible , indeed , we think it is highly probable , that some balances may have to be " aged " for the particular loads they are to carry during the subsequent research .
It is conceivable that under the influence of a given stress , one group of screw 's may manifest , relatively , a large degree of freedom ; the application of a totally different stress may cause the larger degree of freedom to be transposed to another group .
But it is to be hoped that further experiments will show that such tactics on the part of a balance are the exception rather than the rule .
By referring to the numerous papers dealing with atomic weight and other determinations involving very accurate weighting , we find that investigators have frequently made a special point of securing a new balance for their research .
Viewed in the light of the facts we have given , it would appear that this plan might in some cases be productive of uncertain and misleading data .
We believe that the most reliable instrument for highly refined weighings is a well-cared-for balance that has been regularly and freely used for at least twTo or three years .
Under those conditions there will be a tendency for the balance to become aged in a perfectly natural manner .
The author does not commit himself to the extent of asserting that the artificial ageing process advocated by him is thus rendered wholly unnecessary ; on the contrary , he is of opinion that whenever very refined weighings are to be undertaken , the safe plan is to resort to the device for ageing the beam , prior to the inauguration of the research .
In this way alone can it be determined whether or not the balance merits that large degree of confidence which wre at times are apt to repose in it .
Effects Produced by Sudden Changes in the Temperature of a Balance Shelf\#151 ; During the concluding stage of another research , a new and hitherto unobserved disturbing factor recently asserted itself ; this factor assumed the form of a diurnal variation in the R.P. On certain days the values of the reduced R.Ps .
increased very slightly but steadily as the day advanced ; generally , the value attained a maximum by 2 p.m. and then remained , within the limits of experimental error , invariable throughout the afternoon and evening .
The cause of this shifting in the R.P. was not at once apparent ; but ultimately it was suspected to be due to a minute tilting of the shelf upon which the balance rested , the disturbing element of the tilt being in a direction parallel 1912 .
] Temperature Coefficients of a Precision Balance .
599 to the length of the beam .
Any tilt imparted to the shelf will obviously be correspondingly shared by the central pillar which supports the beam ; and as the summit of the pillar will , for a given tilt , describe a larger arc than the base , it follows that ( other conditions remaining constant ) , the pointer will suffer a lateral displacement with regard to the scale placed behind it ; but the actual angular position of the free beam , when referred to a vertical line , will remain unaffected .
That this hypothesis was correct was proved in the following way .
A stout-walled glass tube , having a diameter of 25 mm. , was chosen ; by grinding , one end was adjusted until its plane was very approximately at right angles to the tube 's axis ; the other end was softened and drawn out , so as to form a short , rigid , capillary tube .
A pendulum , consisting of a small brass sphere suspended by a very fine platinum wire , was then introduced , the free end of the wire being passed up through the capillary at the top , and there secured by means of a plug and cement .
Immediately behind the wire , and just above the bob , was fixed a strip of Bristol board , with a thin but well-defined vertical index line drawn upon it .
The distance from the point of suspension to one end of the index line was practically equal to the distance from the central knife-edge of the balance to the lowest extremity of the pointer .
The tube , with its enclosed pendulum , was then set up upon the plate glass slab , upon which the balance stood , and there fastened with a little cement .
For the purpose of measuring the distance between the pendulum wire and the index line , use was made of a reading-telescope , placed at a distance of 2 metres , and fitted with a micrometer eye-piece , capable of indicating 0*01 mm. This apparatus afforded all that was required for detecting and measuring those minute lateral displacements of the balance pointer that might be produced when the shelf was tilted .
As the balance pointer and the tilt-detecting pendulum were of equal length , by measuring the side displacement of the latter we discover the corresponding movement of the former .
In these experiments no attempt was made to measure any lateral displacements that were less than 0*01 mm. ; this was sufficiently small for our purpose , and was therefore accepted as a minor limit .
It may be mentioned that a lateral displacement of the pointer amounting to 0*01 mm. was approximately equivalent to 0*002 mgrm .
Experimenting with the above described apparatus , conclusive proof was soon obtained that the daily growth observed in the R.P. was due to slight vertical movements at one or both ends of the balance shelf .
These vertical movements never equalled 0*2 mm. ; generally , they varied within the limits 0*05 \#151 ; 0*15 mm. Other experiments revealed the cause .
It was found that 600 On the Temperature Coefficients of a Precision Balance .
the disturbances occurred only when the room temperature was varied somewhat suddenly .
When care was taken to slowly vary the temperature , there was no measurable displacement of the pendulum , and reduced R.Ps .
were quite normal .
From the evidence here given in brief , it will probably be gathered that , in the author 's opinion , the balance shelf possessed a temperature lag sufficient under the given conditions to develop a deformation which was manifested in the manner already described .
We were now in a position to control the disturbing element , and the necessary measures for its complete suppression were accordingly taken .
By weighing only after the room temperature had remained nearly constant for some considerable time , all difficulties arising from the cause discussed above have been successfully avoided .
In connection with highly refined weighings , constant use is now made of the tilt indicator .
The instrument is permanently placed upon and near one end of the balance shelf , and from the indications which it offers we at once know whether a weighing may be advantageously proceeded with or not .
Appendix .
Since the publication of my former paper , Mr. L. Southerns , of the Cavendish Laboratory , Cambridge , has drawn my attention to some experiments land observations of his own .
In a paper on the " Dependence of Gravity on Temperature , " * he shows that soon after heating the contents of the calorimeter , which was suspended from one arm of his balance , an apparent increase in weight was noted ; this apparent increase is attributed to a decrease in the buoyancy of the air immediately surrounding the calorimeter .
This view , although correct as far as it goes , is nevertheless incomplete .
The ascending warm air would certainly lengthen the balance arm from which the calorimeter was suspended , and therefore the apparent increase in weight was due to the joint effects produced by the decrease in the density of the enveloping air and the change in the ratio of the arms of the beam .
Such effects are , as I have already shown , quite distinct from those which properly appertain to and are restricted to the temperature coefficient of a balance .
The changes observed by Mr. Southerns could be demonstrated with any precision balance , but with some balances it would be difficult , if not impossible , to render the possession of a temperature coefficient evident .
* 'Roy .
Soc. Proe .
, ' A , vol. 78 , p. 393 .

