]\gt ; of the Earth 's Horizontal Magnetic Force .
375 been that of the great feast of the Niceteria in honour of Minerva 's contest with Neptune for the protectorate of Athens , and the other , the annual celebration of the Marathon victory .
As respects the year 447 , which is one year earlier than the supposed commencement of the Parthenon , it seems appropriate because at that time Pericles would have been supreme , and it is likely that a temple of such sanctity as the Erechtheum would have called for his earliest attention .
It is true thaf the temple remained long unfinished , but of the causes of this delay we are ignorant .
The connection however of the orientation with the feasts above mentioned would have been exactly the same if the date had been 428 .
The apparent discrepance between the orientation date ( as respect the day of the onth ) in the case of the Temples of Jupiter that which is supposed by Mommsen to have been the day of the celebration of the great feast to the supreme god Munychion 19 , the tenth month of the year , which in a general way corresponded with April ) , whereas the orientation dates give for the c , arlier temple March 30-31 , and for the later March 27 , is explained by the possibilities of the Metonic cycle , for when the Attic year as it would in its course on July 11 , the 19th ychion would agree with March 30 , or if July 8 with the 27th of March .
" " Collimator Magnets and the Determination of the Earth 's Horizontal Magnetic Force By C. , Sc. D. , LLD .
, F.B.S. , Superintendent of the Kew Observatory .
Communicated by the KEW OBSERVATORY IITTEE of the loyal ceived May ad June 15 , 1899 .
Sect. 1 .
Introductory .
2 .
Temperature coefficients .
3 .
Induction coefficient .
4 .
Noment of inertia .
5 .
CoefJicient P. 6 .
Tables of mean and extreme values of netic constants .
7 .
Dimensions of magnets , Table III .
8\mdash ; 14 .
Discussion of results of Tables I and II .
15\mdash ; 18 .
Relations veen different magnetic 19\mdash ; 24 .
Probable errors in determinations of horizontal force dne to crrol .
Sirb values of magnetic constants . .
Criticism of formulae for reducing horizontal force observations , from matbematical standpoint .
29 .
Physical sources of , variations of temperature .
30\mdash ; 31 .
, , , , variations of horizontal force and declinatiou .
VOL. LXV .
2 Dr. C. Chree .
C'ollimator Magnets and the Sect. 32 .
Uncertainties in of moment of inertia .
33\mdash ; 34 .
, , , , torsion .
temperaturecoefflcientsinductioncoeffloient. .
39\mdash ; 44 .
Asymmetry in magnets . .
Law of action between magnets .
48\mdash ; 50 .
Concluding remarks .
S1 .
The present paper deals with magnets employed in measuring declination and horizontal force .
More than 100 collimator magnets of English make have been examined at Kew Observatory , and the record of the resuIts forms probably a unique mine of information .
So far as I know , the only use hitherto made of this has been in the compilation of a statistical paper on the mean and extreme values of the temperature and induction coefficients by the late Mr. G. M. Whipple .
The examination of a collimator magnet at Kew Observatory consists mainly in the determination of certain " " constants These are the " " temperature coefficients\ldquo ; and , the " " induction coefficient\ldquo ; and the " " moment of inertia\ldquo ; K. The values found for these " " constants\ldquo ; are utilised in the construction of tables , intended for reducing the observations of horizontal force .
After the tables are constructed one or two observations are made .
Their primary object is to ensure that the application of the leads to satisfactory results , and that there are no instrumental defects j but incidentally they afford the means of determining the magnetic moment , , of the collimator magnet , and also the value of a " " constant , appearing in the expression 2 mm ' for the couple exerted by the collimator magnet on an auxiliary magnet , moment at distance .
The investigations described in the presenf paper have been prosecuted at intervals during the last five years , the pressure of other work allowed .
Their object has been twofold , 1c to find out whether any relationships..exist between the several con stants , and to ascertain where our present knowledge amplification , and where the present tests are least satisfactory .
I shall first explain the real significance of the " " constants and describe briefly the method of determining them . .
Temperature Coefficients.\mdash ; It is assumed that the magnetic moment , , at a temperature of C. , ths magnet being free from external force connected with the moment at C. by the relation ( 1 ) where are absolute constants for the magnet concerned .
If ( 1 ) * ' .
Soc. Proc vol. 26 , pp. 218\mdash ; 222 , 1877 .
naition of the Earth 's ' Horizontal Jfagnetic Force .
377 be taken for granted , and can be determined by observing at any three convenient temperatures .
These temperatures , as the experiment is conducted at Kew Observatory , lie usually within a degree or two of , and C. Magnets , however , destined for Arctic work are exposed to a temperature below C. , while magnets destined for tropical regions are exposed to a temperature over C. The changes of temperature are made rapidly by introducing hotter or colder water into a wooden box containing the collimator magnet .
Changes in the moment of this magnet , aocompanying observed changes in the temperature of the water , are deduced from the variations in azimuth of an auxiliary magnet , freely at a fixed distance from the deflecting magnet .
In a single experiment , the cycle of temperature " " hot " " mean " " cold\ldquo ; is repeated three times , and the mean of the three observations at each temperature is used in the final calculation .
It is customary to have two completely independent experiments on different days , and to take the arithmetic mean of the values deduced for and on the two occasions .
The exact times are noted at which the several readings are taken , and suitable corrections are applied from the readings of the magnetic curves for variations in the horizontal force and declination .
S3 .
Induction Coefficient.\mdash ; This is denoted by , and really means the temporary change in the magnetic moment of the collimator magnet due to unit change in the field ( 'parallel to the magnet 's length ) , it being assumed that the relation between temporary moment and strength of field is linear .
The experiment consists in observing the angles through which an auxiliary magnet is deflected out of the magnetic meridian by the collimator magnet , the latter being vertical , with its north pole alternately up and down .
The vertical plane through the centres of the deflecting and deflected magnets is perpendicular to the latter 's axis , but the centres of the magnets are not in the same horizontal plane .
The change in the inducing field being double the intensity of the vertical force at Kew is nearly C.G.S. unit .
As a rule , only one complete experiment is made , but this involves inverting and reinverting the magnet several times .
Originally was measured in British units , so that conversion C.G.S. units was necessary in many cases .
S4 .
Moment of Inertia.\mdash ; This means the moment of inertia of the magnet and all its appendages , when at a temperature of C. , about the suspending fibre .
A collimator magnet is a hollow steel cylinder , about cm .
long and 1 cm .
in external diameter , with screws cut on its inner surface at both ends .
The appendages consist of two small cells , one holding a lens the other a glass scale , screwed into the * A special apparatus is employed whose descriptiou would occupy undue spaoe .
The methol is practically described on pp. 151-3 of Lamont 's ' Erdmagnetismus .
' Dr. C. Chree .
ColTimator Magnets and ends of the magnet , and of a brass stirrup arrangement which the magnet and affords the means of supporting , parallel to it , an auxiliary solid brass cylinder .
This brass cylinder is a regular geometrical object , whose moment of inertia can be calculated from its weight , length , and diameter .
The actual inertia experiment consists in observing the timss of vibration of the magnet , under the horizontal force , when the auxiliary bar is in the stirrup , and when it is removed .
To reduce the possible effects of variation in force or temperature , four complete series of vibrations are made , the first and fourth without , the second and third with the auxihary cylinder .
Allowance is made for the departure of the mean temperature from C. It is customary to make two independent determinations of the moment of inertia , usually on different days , and in the event of serious dispancy a third experiment is made .
In all the older experiments conversion from British to C.G.S. units was necessary .
S5 .
CoefficiemX P.\mdash ; The meaning of this has been already explained generally .
I need only add that in the deflection experiment the axes of the two magnets are perpendicular , and the centre of the deflected or , it is called , " " mirror\ldquo ; magnet lies on the axis produced of the collimator magnet .
The general expression for the couple exerted such a symmetrical position , the centres of the magnets being at distance , is where , .
are constants , whose values depend on both the deflecting and deflected magnets .
As increases , the terms involving the higher negative powers of tsnd to vanish relative to the first term ; and in the presence case it is assumed that the term involving is the last that need be retained .
When this is true we can determine by comparing the couples answering to any two different values of .
The distances originally adopted at Kew Observatory when British units were employed were 1 foot and feet ; the distances now in use cm .
and 40 cm .
For several reasons , I recorded the value not of but of at 30 cm .
, allowing in cases where Briti\amp ; units had been used for the small difference between 30 cm .
and 1 foot .
S6 .
The number of makers of collimator magnets is not large , and the differences between the patterns ars mostly small .
Still it seemed undesirable to wholly disregard the differences that unquestionably exist .
Accordingly , in summarizing the resuIts contained in the Observatory records , I have divided the magnets into six groups , distinguished by the letters A to F. Four of the groups , , and , contain magnets from one maker only , and are presumably fairly homogeneous .
Group is the most containing magnets by thres if not four makers ; group contains magnets from two makers only .
Determi , of the Earth 's Horizontal agnetic Force .
379 Table I gives the mean values of the various quantities specified for the several groups ; while Table II gives the extreme values .
Table I.\mdash ; Mean Values .
Table II.\mdash ; Extreme Values .
Data were wanting for in the case of eighteen out of the 130 collimamagnets , and there were no data for in the case of three magnets , group C. In these twenty-one cases was calculated from the Dr. C. Chree .
Magnets and the recorded value of the angle through the magnet deflected auxiliary magnet at a distance of foot , that being .quantity determined in the induction experiment .
Old magnets , when remagnetized and retested , have been treated as new magnets and counted separately .
S7 .
The discussion of Tables I and II would have been much simplified if one had possessed a record of the weight and dimensions of the collimator magriets themselves .
It is not , however , customary to these quantities at the Observatoliy , as they in no way enter in:to the tables used in reducing the observations .
The only data at my disposal , bearing directly on these points , are results of special measurements recently made on six magnets , three belonging to group , and three to group E. The results were as follows , the numbering being pnrely arbitrary the length are in centimetres , the eights in grams:\mdash ; Table III.\mdash ; Dimensions and Weights of specimen Magnets .
iameter a The specific gravities found for two of the magnets were approximately equal , the mean being From this we conclude that in both groups of magnets the wall thickness is less at the extreme ends , where an internal screw is cut , than elsewhere ; the mean internal diameter deduced from the data being cm .
for the magnets of group , and cm .
for those of group E. The actual volumes of steel in the two cases were respectively and There has been but little variability in the length or external diameter of the magnets of any one group ; the volume however , if we may judge by Table IIT , is somewhat more variable .
We should have more exact information on the point if meant the moment of the magnet alone , but as matters stand , our conclusions are exposed to some uncertainty .
The appendages , as we see from Table III , .
weigh about as much as the magnets themsel , their main mass lies however nearer to the axis of rotation .
In fact , according to the measurements Deterrnination of the Earth 's Eorizontal llfagnetic Force .
381 made on the special magnets the appendages contribute only from 25 to 30 per cent. of the value of K. Again , the size of the appendages is determined main by that of auxiliary brass bar , which according to the Observatory records is nearly constant for magnets of the same make .
It is thus improbable that variations in the appendages exercise a large inflnence on the values of in the several groups .
S8 .
Taking all the data into oonsideration , I conclude from Table I , that the magnets of group A are on the whole distinctly the largest , and those of group very distinctly the smallest ; that the other four groups are very fairly similar in this respeot , though the aYerage magnet of group is almost certainly a t1ifle smaller than that of groups and C. The variability in size was really one of my principal reasons for grouping the magnets .
The values of and should increase , , with the size of the magnets , and all analysis which overlooked the difference between groups A and might well prove misleading .
So far as individual groups are concerned , I do not think that the neglect of possible variations in the size of the magnets is likely to prove serious .
My reasons for this opinion are partly based on phenomena dhscussed later in SS15 and 16 .
S9 .
Before proceedi11g further , a second source of uncertai1lty must be noticed .
The moment of the residual magnetism in a magnet not determined solely by the magnetic quality of the steel , it depends on the conditions under which magnetisation took place , on the time elapsed since that event , and on the usage to which the magnet has been exposed .
In the present case I have strong reasons to think that the very large maiority at least of the magnets were magnetised under a unlform set of conditions .
Until quite lately it was the lnvariable practice to magnetise all the magnets at the servatory itself , and the same coil with similar battery power has been use for the last forty years .
The magnet is not placed inside the coil , but is stroked in a uniform way on the end of a very massive slightly proiecting lron core .
The capacity of this core to hold out a heavy pole piece has been regarded as a caiterion of the battery being in proper order ; and as apparent " " saturatio1l\ldquo ; is reached in an ordinary collimator magnet with the below standard condition , the test , though a rough one , appears fairly satisfactory .
There is a variety of indirect evidence as to the uniformity of the conditions .
For instance , the dates of magnetisation of the ten magnets of largest in group , arranged in descending order of magnitude , were as follows : , 66 , 79 , 73 , 95 , 84 , 83 , 94 , 75 , 89 , 68 .
As regards possible loss of magnetic moment , the magnets , with one or two probable exceptions , had all been magnetised but a short time before and the various ' constants\ldquo ; were deternuned .
382 Dr. C. Chree .
Coltimator Magnets and the S10 .
Magmtic JIoment .
Taking into account the variability in size which we have decided to exist , it would appear from Table I that the capacity for assuming permanent magnetism has been nearly the same in the four groups , and ; but the magnets of group have been distinctly less receptive than the others .
It is difficult to say exactly how the magnets of group stand .
, They are clearly much below the average size , and their mean is at least as large compared to the means in the other groups as one would expect from a comparison of the mean moments of inertia .
Table II shows the variability of to have been much greater in each group than that of K. The great variability in is noteworthy , because when is small corrections for torsion , temporary induction , \amp ; c. , increase in relative importance .
It has been the practice of late years to reject magnets showing exceptionally small magnetic moment , but the data from such reiected magnets are not included in Tables I and II .
S 11 .
Temperature Coefficienfs.\mdash ; Out of the whole 130 magnets was not a single case in which the ithmetic mean of the values found experimentally for was negative .
Thus within the experimental limits , say to C. , the rate of change of magnetic moment with temperature invariably increased with the temperature .
The mean value for in groups , and , is practically identical , and is very distinctly larger than the mean values for groups and D. Group is specially remarkable for the smallness of the temperature coefficients ; in fact , the largest value of found in a magnet of this group is very little greater than the mean value in groups and F. The mean value of is fairly similar for the different groups , but groups A and have somewhat smaller values than the others .
The variability of both and is , as we see from Table II , very considerable in all the groups except , where varies but little .
The term is in general small compared to the term , and consequently the probable error in the determination of is large .
The variability of may for this reason be somewhat exaggerated in Table II .
S12 .
Induction Coefficient.\mdash ; As explained above , is the product of the volume of the magnet into its permeability , as determined by 1eversing a field of about C.G.S. unit , the magnet being possessed of a large permanent magnetic moment .
Table I shows that the size of the mean in the first five groups has little relation to the size of the mean .
This is significant , because and should vary in the same way with the volume of the magnet .
At the same time , the exceptionally small volume undoubtedly possessed by the magnets of group is almost certainly the cause of the small found in that group ; and we shall probably be correct in nination of the 's Horizontal Ilagnetic Force .
383 concluding that the larger mean possessed by group A as compared to groups , or , is due to the possession of a larger vo ume yt lume th first-mentioned group .
The pre-enunence of the mean group rou Ei however , explicable only by supposing that the magnets of that group possess a considerably larger permeability than the others ; an this larger permeability , as we see from Table II , is a property not of one or two magnets , but of the whole group .
On the whole , as Table II shows , is much less variable in the groups than or The greatest recorded value of , viz. , , occurs group D. I feel , however , some doubts respecting this and a second large va lue , , iound for a second magnet of the same group .
The two magnets for which these values are recorded were tested in 1864 ; and one of them , when retested many years afterwards , had a less than 7 .
If we excluded these two magnets , we should find for group a mean much the same as that of group C. S13 .
Coefficient P.\mdash ; In about half the cases was determined from only one observation , and under such circumstances the probable error is considerable .
Still I do not think that the mean values given in Table I can be much in error .
It will be noticed that no magnet of group gave a positive , and that no magnet of groups or gave a negative P. In groups , and D. ' positive and negative values occur in fairly similar proportions ; but investigation showed a great predominance of negative values amongst the older magnets of group , while in group no positive value appeared in the six oldest magnets .
As pointed out in S 5 , depends on the deflected or mirror magnet as well as on the collimator magnet , and I am inclined to ascribe the interesting difference between older and newer unifilars mainly to change in the pattern of the mirror magnets ( see S46later ) .
As appears from either Table I or Table II , when negat ve usually numerically smaller than when positive .
When an observer employs a unifilar strange to him the probable error in is doubtless than when the instrument is one to which he is thoroughly , and it would not be unreasonable to attribute to this cause some of the large values recorded in Table II .
It happens , however , that three of the largest values of in group were each based on three independent observations .
resuIts of the individual experiments were as follows , the numbering being purely arbitrary:\mdash ; 384 Dr. C. Chree .
Collimator Magnets and the The probable errors in the mean values found in these three cases are certainly sensible , but still they form only a comparatively small percentage of the value of P. Certain drawbaoks attending large values are deaIt with later .
S14 .
A general survey of Tables I and II shows merits and demerits in most of the groups of magnets .
Thus in group the temperature coefficients are exceptionally small , but on the other hand the magnets are somewhat weak .
Again , in group the magnets are exceptionally uniform in quality , but they possess exceptionally large permeability for temporary magnetism .
Relahonships between Magnetic Constants .
S15 .
In attempting to determine the existence or non-existence of .
relationships between the magnitudes of the several magnetic constants , one naturally turns first to the large group C. To determine whether the size of the permanent magnetic moment influenced the other quantities , I arranged the eighty-two magnets of the group in sub-groups as below , and found the mean values of the several constants for each .
Table \mdash ; Analysis of Magnets of Group according to Magnetic Moment .
It is only natural to expect a sensible departure between the means of any single property in a sub-group of ten magnets and in a whole group of eighty-two magnets , supposing the sub-group selected by pure chance .
Bearing this in mind , we must , I think , conclude that no clear relationship exists between the value of and that of , or The absence of apparent connection between and seems strong evidence of the comparatively small variability in the actual size of the magnets .
The fact that is nearly , if not quite , independent of is important , because it is expedient eteris paribus that should be as large as possible .
If , or had shown an appreciable tendency to vary with we should have been led to suspect , as a priori probable , tendency to alteration in these constants as the magnet grew weaker with age .
The absence of any apparent connection of the kind is not , however , proof positive that no such tendency exists ( see S37later ) .
jtermi , of the Earth 's Eorizon.tul Maynctic Force .
385 S16 .
I next arranged the eighty-two mag , nets of group in subgroups , according.to the value of , with the following results:\mdash ; Table \mdash ; Analysis of Magnets of Group according to Size of Induction Coefficient .
There is here no trace of a connection between and , which supports the conclusion drawn from the absence of apparent connection between and K. There is probably a slight tendency when is very decidedly below the mean for to be slightly .
but the tendenoy in to increase as diminishes is conspicuous .
The large value for , when negative , lies between 5 and 4 has no real sigIlificance , the mean being based on only two magnets .
The one clear and important relationship brought out by Table is between and .
Large values of these two constants uIlquestionably have a tendency to occur together in the magnets of group C. Out of the twenty-six magnets whose exceeded 6 , no. less than eighteen had a above the mean ; while of the thirteen magnets whose fell short of 5 , only two had a above the mean .
As the result seems important , I submit the following analysis , showing the distribution of the different values of in group Table \mdash ; \mdash ; All values . . . . . . . . . . . .
4 17 19 13 7 8 1 2 From 7 . . . . . . . . . .
2 3 3 6 4 5 1 2 6 , , . . . . . . . .
2 3 3 7 5 2 2 , , , 5. . . . . . . . . .
2 3 6 5 2 1 5 , , 4. . . . . . . . . .
3 5 4 1 The table is to be read thus : Of the eighty-two magncts , four had a value of between 150 and 200 , and of these , two had a lying between 6 and , while two had a lying between and 5 .
Table VI certainly supports the conclusion drawn from Table As a further check on this conclusion , I arranged the eighty-two magnets in two sub-groups , according as was above or below the mean , with the following result:\mdash ; The mean of the first sub-group of Table is distinctly larger than that of the second sub-group , but the evidence of a tendency in large values of and to go together is not wholly conclusive .
Thus out of the thirty-three magnets whose exceeded the mean , fourteen had a below the mean ; and two of these fourteen had the largest 's of the group .
The force of such notable exceptions is weakened , however , by the consideration that an experiment which makes slightly too big is more likely than not to make considerably too small .
S18 .
Groups and are the only ones beside which are sufficiently numerous and homogeneous to merit analysis .
In the two following tables I give the results obtained by subdividing each of these groups into two numerically equal sub-groups , according to the size , first of , second of Table VIII.\mdash ; Mean Values of Constants in Sub-groups .
Determination of the Earth 's Borizontal Magnetic Force .
.387 Table IX.\mdash ; Mean Values of Constants in Sub-groups .
Values of were wanting for three of the twelve magnets of group B. In one respect Tables VIII and IX unquestionably agree with the results established for group ; they show a distinct tendency in to be large when is large .
There is , however , in Table IX no association of large values of and .
In the case of group there is apparent association of large values of both and with large values of , suggesting that the size of the magnets has exerted a slight sensible influence .
In the case of group the differences between the various magnets are so extremely small that we could hardly hope to detect any relationship that was not very intimate and potent .
The previous part of the paper has been mainly historical and descriptive ; it remains to consider the subject from a more critical standpoint .
Probabfe Errors in Determination of yrizontal Force due to in Values of " " \ldquo ; S19 .
Let us first assume that the methods of determining the " " constants\ldquo ; and the formulae employed in calculating the horizontal force are alike abovs suspicion , and investigate on this hypothcsis the llncertainties introduced by the probable errors in the values found for the several ' ' constants To do this , we mnst consider the formulaeThe notation not already explained is as follows:\mdash ; X horizontal component of magnetic force ; semi-period of vibration , corrections having been applied , if necessary , for rate of chronometer and for finite arc of vibration ; torsion couple , during vibration experiment , when torsion angle unity ; deflection angle , during deflection experiment , when distance apart of magnets ' centres ; temperature during vibration experiment , , , , deflection ( 5 ) .
From ( 5 ) we see noteworthy differences in the consequences of errors in the different " " constants In the oase of it is not the absolute size of the error that counts , but the ratio it bears to the size of ; while in the case of , and the temperature coefficients it is the absolute size of the error that counts .
In all cases increases with X , so that the obsolute effect of a given error in any one of the " " constants\ldquo ; is greater where the horizontal force is large than where it is small .
S20 .
In estimating the probable errors in the several constants , I have confined my attention to the cases in which the accepted results were based on two , and only two , experiments .
If be the difference between the values given by two experiments for a certain quantity , the probable error in the arithmetic mean , , of the two determinations is In many instances the value of a " " constant\ldquo ; was based on only one experiment .
In such cases we may reasonably assume that the single experiment was , on the average , neither better nor worse than the *Cf .
Stelvart and Gee 's ' ] ' vol. 2 , pp. 298 , allowing for difference of notation .
Determination of the Barth 's Horizontal Magnetic Force .
389 average experiment which formed one of the coupls usually made .
This would givs for the probable error in the value of a " " constant\ldquo ; based on one experiment only , the value where is the arithmetic mean of the probable errors found in the cases where two experiments were made .
In the few cases where constants were based on three independent experiments , one could , of course , have calculated the probable errors , and found their mean .
The number of such cases appeared , however , too small to give a satisfactory mean value .
S21 .
Taking first the moment of inertia , I examined seventy-one cases in which two independent determinations had been made .
Employing to represent the difference between the two observed values , whose arithmetic mean is , I found Mean The corresponding mean probable error is given by and answering to this the probable error in X is Ascribing to X the value C.G.S. which is not far from the present mean value in Great Britain\mdash ; we should have SX , approx. Taking the same value for X , I also determined the law of incidence of the probable error in the seventy-one cases examined .
The results were as follows:\mdash ; Table X. Prob. error ) between Number of magnets 18 8 1314468 As measurements of horizontal force are usually taken to C.G.S. unit , we see that error in the moment of in ertia- may be expected to affect the last significant figure , in these latitudes , in fifty- three cases out of seventy-one , or practically in two cases out of three .
In one-1linth of the total number of cases the able error in X shown by Table X reached or exceeded five units in the last significant figure .
Dr. C. Chree .
Coliimator Magnets and the Also it must be remembered that in equatorial regions X may be double the value assumed in Table X , and that in this case varies as X. S22 .
The number of instances where two observations had been made of were I examined forty-one of these in all .
Representing by and the mean of the semi-differences between the two observed values and the mean probable error respectively , I found The corresponding probable error in X , ' as 30 cm .
, is , irrespective of sign , , approx. This is troublesome to deal with ; because depends both on X and on the magnetic moment .
As a first approximation we have in fact To get an idea of the probable error arising from error in , suppose X , as at Kew , and , as in the average new collimator magnet .
This gives , approx. , approx. Thus when there are two observations of the probable error will in the average case fail to affect the last significant figure , supposing X measured as usual to C.G.S. Of course in a good many individual cases the probable error in determined from two observations , was sufficient to affect the last significant figure at Kew .
More often than not has been derived from a single experiment , and in the maiority of such cases we should conclude that the probable error was large enough to affect the last significant figure in X , measured at Kew .
Owing to the occurrence of a term in the expression for , we see that where X is large , and is small , error in may be very much more serious than in the case we have selected for numerical treatment .
S23 .
The influence of errors in and on X is difficult to present clearly , as it depends both on the difference and the mean of the temperatures and existing during the vibration and deflection experiments .
If and are equal , it does not matter\mdash ; the fumdaof the } ' .
391 mental formulae being granted\mdash ; how largely are error : ) this is an exceptional occurrence , especially field observations .
To consider all possible mean temperatures seemed mnecessary , and I thus confined my attention to the three cases C. , C. ; corresponding to which , .
In other words , I found the differen ce between the ) of the three quantities and , treated independently , and the corresponding three independent probable errors .
The results , derived from seventy magnets , were as follows:\mdash ; Mean semi-diffeQuanti } rence bIeall probable error Corresponding to this , we have for the meaIl pahle errors in X Mean temperature C. C. C. The probable error is conspicuously less near the middle part of the temperature range covered by the actual experiment ohan near either limit of this range ; and this is only what we should anticipate .
Vhen the mean temperature during a horizontal force observation , at Ke is C. , it would in the average unifilar require a difference of fully between the mean temperatures during the vibration and deflection experiments to make the probable error in and affect the fifth significant figure in X. So large a temperature difference as this need hardly ever be feared in a fixed observatory .
The result is so far comforting , but does not justify the conclnsion that error in the temperature coefficients is a wholly improbable of error in X. In some individual cases the probable errors for and were five or six times larger than the mean .
Agaill , in a considerable number of instances , and have been derived from only one expeI .
Finally it should be noticed that the probable error increases with X , and that on the whole X is largest in equatorial regions where the temperature is high , and consequently errors of eriven magnitllde in and most effectivs .
S24 .
The two terms and - were included with the object of showing how errors in the values assigned to or to the torsion affect X. I have , however , no satisfactory data as to the size of the probable errors in or the torsion coefficient mder normal conditions .
The torsion coefficient varies from thread to thread , and also with the dampness of the air .
It is in fact )-ariable , and is nsually VOL. LXV .
2 Each of these results is ou a large ) of observations , I shonld hcsitate to that error plays no part in the rent fctnations .
There are of course frequently minor magnetic ) ances d force observations , and two or three ding v of sensibly aflect a mean though from ) ations .
the annnal mean has invariably made tive , posive hles from observations are by means lmconunon .
If co1lsider that the in the above table are based ettch on lnmerous iments , takellat a fixed station under the best conditions , we mllst , I think allow Ulcertainty in the correction is very ollrce of in S .
If be the by the axis of the collimator magnet ith the nlagnctic meridian at time , the equation of motion ( 6 ) , Determination of Earth 's Horizontal Forcc .
where is the moment of inertia of the magnet and appendages , and the induction coefficient at temperature If be kept small enough this gives an isochronous vibration whose half period is given by whence .
Unless be kept very small , a correction becomes necessary for " " finite arc\ldquo ; of vibration ; and we then encounter the difficulty that the torsion coupls is and not This is rarely , however , of practical importance , except at places herc X is specially small , supposing one avoids coarse suspension f ) At Kew very seldom reaches , and need never excee and under such conditions may be freely written for .
The correction for " " finite arc\ldquo ; then presents no peculiarity .
A second criticism that may be passedis that ( 6 ) makes no for air resistance ; in the absence of experimental data I to say on this point .
S 26 .
In the deflection experiment the deflecting and deflected magnets are at right angles , the latter making an angle 21 lvith magnetic meridian .
Supposing the temperattlre , the induction coefficient of the coilimator magnet during this experiment , and X ' the holizontal force , have\mdash ; X ' . .
( 8 ) , assuming the term negligible .
Here is the actual distance at temperature between the centres cf the two magnets .
Unless a large magnetic storm is in progress\mdash ; in which case an lute observation of horizontal force is worthless\mdash ; we may regard X and X ' as equal in the snlall terms of ( 7 ) and ( 8 ) , and so these equations as X ' .
Thence , eliminating ?
, we have This diflers from ( 4 ) in scveral respects ; some of these little real significance .
Dr. C. Chree .
uimator Magnets the Thus we have in ( 9 ) as against in ( 4 ) ; but this only means that the X appearing in ( 4 ) is in reality a mean between the values possessed by the horizontal force during the vibration and deflection experiments .
In actually comparing the result of an absolute horizontal force observation with the magnetogram one measures the curve ordinates at the mean times of the two experlments\mdash ; times separated usually by thirty or forty minutes\mdash ; and takes the mean of the two ordinates as corresponding to the X deduced from ( 4 ) .
Again , the appearing in ( 4 ) is really taken from a table which allows for variations of temperature , and the same is true of S27 .
The first differenoe of real significance is that ( 9 ) contains , while ( 4 ) , on the other hand , has .
Supposing X measured as usual to five significant figures , this becomes obiectionable when the value of X is affected by so much as C.G.S. unit .
The limiting value of for which the substitution is justifiable is given by or , approximately , This gives\mdash ; For X , approx. , The mean value found for , when is positive , in Table I is just in excess of , and values in excess of are very common .
Thus the employment of in ( 4 ) is , to say the least of it , frequently uniustifiable .
As regards the possible size to which the error in question might attain , we should have in the case of the largest given in Table II for X , error , , , S28 .
The second difference of note between ( 4 ) and ( 9 ) is that in the former as against in the latter .
Overlooking for the present a possible difference between and we see that obiection arises when ceases to be negligible .
So far as the term in is concerned there is no cause for apprehension , as would be an exceptionally value for nination of 's Borizontal Magnetic Forc The case of the temperature terms is less satisfactory .
Taking , for example , the mean values of Table I , we have Following exactly the same procedure as in the case of thence conclude that the treatment applied to the temperature terms will on the average begin to introduce error , supposing X 0.18 when reaches C. When a magnet with large temperature coefficients is used in hot weather at a place where X is large , the error due to the treatment of the temperature terms becomes very sensible .
Suppose , for instance , values exceeded in one actual case , then we have for This gives error of approximately C.G.S. unit X S29 .
Under this heading I shall discuss certain grounds of uncertainty which may affect the accuracy of the calculated valuesl of the horizontal force .
The considerations are mainly , but not us physical .
An objection to both the un amen a orm 1 he fundamental formulae ( 7 ) and ( 8 ) is that they represent all the quantities involved as co11stant , whereas genera magnetic force and declination , temperature , and moment magnet , are constantly altering .
dIn reality , however , the appearing in ( 7 ) is the mean of two rea ings of a thermometer adiacent to the collimator magnet , taken the one immediately before , the other immediately after , the vibration experiment ; while the appearing in ( 8 ) is the mean eg ei ht thermometer readings taken nearly lsly with the dual observations of the deflection angles .
Thers is thus very fair provision for changes of temperature so long as the temperatnre changes slow continuously in one direction , provided the thermome ers rea ters real record the temperature of the magnet .
In a fixed observatory these conditions are probably as a rule fairly secured if the magnet an ermo th meter be freely exposed to the air for five or ten mlllutes .
th experiment commences .
*November 22.\mdash ; Compare a paper by Wild ( which I had overlooked ) , ' strial Magnetism , ' , 1897 , pp. irregular variations of horizontal force and declination is in all probtbility a very complicated problem .
S31 .
The possibility of satisfactory horizontal force measurements depends largely on avoiding times of serious magnetic ( hsturbance .
shorter the time occupied by the observations , the better the chance of accomplishing this .
Supposing a uniform procedure adopted , the time occupied by a vibration experiment is independent of the observer ; it is shorter the stronger the magnet and the greater the horizontal force .
It is thus of especial taIlce that collimator magnets intended for work in arctic regions , where the force is low disturbances large and , should be of high magnetic moment .
In the deflection experiment a great deal depends on the of the observer , and on the make of the unifilar ; one observer may do in twenty minutes what occupies another for fifty .
Extreme conscienmay be a positive demerit , owing to the excessive time spent in adjustments and readings .
A variety of other criticisms will be considered which apply to some one physical quantity , and are classffied accordingly .
S32 .
of inertia.\mdash ; In defauIt of any simple means of directly measuring the variation with temperature of the moment of inertia , this variation is calculated by assuming for the coefficients of linear expansion : in the auxiliary brass bar , , , , collimator magnet .
is of course the uncertainty that the assumed mean values may of the not apply very exactly to samples of ) stccl ployed in any individual case ; and there is the fnrther ) the coefficient is applied to a composite nloment of inertia of which 25 to 30 per cent. is contril)uted by the which are mainly brass .
The expansion coeffi(\ldquo ; ients are in deducing the value of at C. from the found some higher temperature , and then in calculating from the at thus , a table giving at all likely to occur 1 ] actnal use .
Any error that may .
trise in this will be rger the more remote the temperature from that isting dnring the special inertia experiments .
Its probable crnitnde c er , be directly arrived at in the absence of of the of specimen magnets , their tppend and auxiliary A second source of uncertainty is the that is to assume miform density in the auxiliary ) S33 .
Torsion of threa\mdash ; This is required ( 1 ) in reducing the declination experiment , ( 2 ) in the ) tion c dete mining the horizontal force .
a considerable time prior to a declination Kc the silk fibre is stretched ) hanging from it lletic phunmet in weight to the its .
As the mag1letic meridian is very pproximately , it is to that when the plunlmet replaced ) the tiou mgnet there shall be very little twisting of the may thns , at reasonable ) ability , regard the thread at the of the meant as practically free from torsion .
After the experimcut is conclucied , the magnet is again replaced by the lmmet , some hours its position is noted .
In this we cdll tell approxinlatelv the angle through which the lower end of the has tnrned since the last declination reading was taken .
It is bnmed that represents the of torsion in the silk when the cnded , that this torsion was introduced gradually in the of the expel'imeut .
As the magnet has to bc varionsly tted , there is ntecedent mprolJability in the hypothesis ; but it , I think , be that the taking the mean torsion angle dluing the iment is itrl emendation whose success is likely to be variable .
To allow for of torsion , one introduces givcn into the and notes the consequent change of azimnth in the magnet .
The procedure adopted at , is to tnrn the } ) head carrying the thread throngh first in one direction then in another , noting the corresponding equilil ) positious of the llet .
S34 .
The occasion for torsional correction in the ) meant is the existence of the couple in ( 6 ) , the term in ( 7 ) .
The value of is ( from the ) crved in the azimuth of the collimator } ctn.hcll the ) When one uses a suspension sufficient for the collimator mngnet alone , without the auxiliary bar , one oan with an average magnet get as low as at Kew .
Now an error of 10 per cent. in through neglecting the variation of with temperature or the fluctuations of X at a fixed station is quite out of the question .
On the other hand , suspensions such as are frequently used to carry the auxiliary bar as well as the magnet , and ars intended to stand a good deal of rough handling , make at Kew as large as .
In such a case error may occasionally arise not discriminating between and In practice the most probable sources of error are inaccuracy in the torsion experiment and ariation in , owing to variation of moisture between the vibration and torsion experiments .
S35 .
\mdash ; In the method of determining temerature coefficients in vogue at Kew the collimator magnet is fixed lnside a ooden box , rigidly attached to the which carries the nnifilar .
The calculation assumes the mirror which is suspended as in the deflection experiment\mdash ; to be exactly perpendicular to the collimator but , ring to the position of the latter being fixed , this is in general only approximately true .
Suppose the deflected magnet to make an angle the magnetic meridian , and an angle lnstead of the collimator magnet .
The deflecting force is in tcrnls of X by the equation lination of Barth 's Magnetic } If alters owing to ohange of temperature in the collimator magnet , its increment and the increments in and ?
are connected the relation .
But S and are equal and so , supposing small , have The assumption , taoitly made in practioe , that the magnetic axes perpendioular is legitimate only when we may neglect The size of the error varies with the moment of the deflecting magnet , and also with the secular change of deelination .
A slight change is , I think , desirable either in the experimental arrangements or in the calculation .
S36 .
A more strictly physical objection to the temperature experiment is that in it the changes of temperature are very sudden , whereas in actual use they are gradual .
I am not of any experiments bearing directly on the question whether a change of from C. to , occupying only a minute or two , has the same ?
effect on a magnet as an equal change occupying as many hours .
Theoretically it would be very desirable to make the changes of temperature but in practice this is troublesome , owing to the disturbing influence of changes in declination and horizontal force .
A change of C. in the temperature of a oollimator magnet at Kew seldom alters the azimuth of the deflected magnet by more than , so that it is important to avoid the risk of sensible disturbances so far as possible , even corrections are apphed from the magnetic curves .
Another objection is that sudden changes of temperature are apt , like mechanical shocks , to permanently diminish the magnetio moment .
It is clear that unless suoh permanent loss is small , satisfactory of and are unlikely to result from a oalculation whioh lmes that the temperature effects are purely temporary .
In the normal temperature experiment at Kew the cycle " " hot " " mean " " cold\ldquo ; is repeated three times , and the arithmetic mean of the three readings of the deflected magnet 's azimuth , to the three " " hots\ldquo ; is attribtlted to the temperature whioh is the arithmetic mean of the three observed temperatures .
This method probably eliminates to a extent the influencs of permanent supposing them to exist , but it also renders their tion somewhat difficult .
To investigate the matter I took a ) of ture experiments , and employing the values for and * November 7.\mdash ; Since this writteu the necessary ange in the apparatus has been effected .
Magnets and the the of the gnetic moment , to a common temperature , at each of the three " " hot " " mean and ' ' cold\ldquo ; xeadings .
, for instance , if the three " " hot\ldquo ; temperatures were , and , the nagIletic moments were reduced to the mean temperaiure 3 The departures from the mean temperatures in this case are so small that trifliug errors or are immaterial .
Corrections were careapplied in all cases from the magnetog for changes in the horizolltal force and dechnation .
In all , sixteen experiments made on ten different magnets with ; the mean results are given in Table XII .
By is meant the diminution found in the magnetic moment divided by the original value of the moment at atmospheric temperature .
This is multiplied to avoid decimals .
Table XII .
Iean Values of nperatule .
2nd .
3rd .
3rd reading-lst .
" " Hot\ldquo ; 266 95 383 Iean " " 260 346 " " Cold\ldquo ; 166 56 329 Ieall of 3 stages 319 In one case ondy one cycle taken .
cannot accurately to one in 100,000 , much to one part in a million , so that a high degree of accuracy can hardly be claimed for the above figures .
It seems , however , perfectly clear that ormally there is a loss of magnetic moment , but that at the same time the average loss is very small .
These conclusions were supported by a nmber of other cases which I examined , though less carefully .
On the average , according to Table XII , a complete temperature experiment causes a loss of per cent. in the value of , and ( of this loss much the greater part occurs during the first temperature cycle .
The true reversible temperature variation of ring a cycle verages very sensibly over 1 per cent. thus the shock effect can cause but little error in the values of and in the average magnet .
Cases occur , however , in which the loss of magnetic moment is very ( considerable , and the values of and may have suffered in conse .
The property seems to depend on the tempering , not on the chemical character of the steel ; for various magnets which have been rejected for defective retentiveness have behaved normally after further treatment by the maker .
In some of the experiments summarized in Table XII there was a * It is intended in future to have four temperature cycles , discarding the results of the fir cycle .
of Earth 's Magnetic 401 of moment at one or more of the stages .
In slight apparent lncrease fact , in an old magnet not recently remagnet se , an a ] and in a luagllet hich had been put through a previous temperature experiment on 3 three days before , the moment appeared greater on the thir occulrence of the " " hot\ldquo ; temperature than on the first .
In investigating these permanent changes of magnetic moment I received valuable assistance from .
R. S. ]yhipple , then assistant at Observatory .
S37 .
There are several other sources of umcertainty affecring temperature coefficients .
For instance , a small fall of tempsrature may not have an effect exactly equal and opposite to that of an equal rise .
Again , it is conceivable that the va ues lues of and may depend to some extent on the age or strength of the magnet , an so may a so ma in of time .
On this point I have some interesting data .
In 1894 three collimator magnets after long absence in India were sent to Kew Observatory to be reported on .
The opportunlty was taken to redetermine their temperatllre 11 determined in 1865\mdash ; before sending them for repair to the maker .
After their remrn they were remagnetised and tested as gives the valtles deduced for at , and C. , employing the foumd for Table gives particulars as to the values of the moments of these magnets at the times of the temperature experlnlents , and the oss loss of moment which occurred between 1865 and 1894 .
The llumbsring of the magnets is arbitrary , but is the same in both tables .
Magnet ii discarded in 1894 , being slightly chipped .
It will be seen that the agreement between the mean tcmperature effects in 1865 and 1894 could rdly be better , notwithstalldlng the large dimilntion in the magnetic moments .
On tlJe other hand , the values found for and in 1895\mdash ; values based in each case on two consistent experiments\mdash ; are unmistakably higher than those found in 1865 and 1894 .
There is no apparent reason for this except the fact that the magnets had been remagnetised .
The makers , in reply to inquiries , stated explicitly that " " the magnets were only cleaned seem driven to the conclusion that whilst a gradual loss of magnetic moment not appreoiably influence the values of the temperacoefficients , the remagnetisation of a magnet does influence these coefficients , in at least some cases .
S38 .
Indnctio method of determining has been already described in S3 .
There are here also several uncertainties .
The assumption that the relation between the temporary moment and the field is strictly linear may not be stlfficiently exact ; C.G.S. is a considerably stronger field than Lord Raylsigh found to limit the linear part of his induction curves , and recent German observers seem disposed to narrow his limits .
The accuracy of the assumption is doubtless less in some oollimators than in others , but I have no data on which to proceed .
*Again , it is open to whether a magnet possessed of a large permanent moment responds equally to the action of two small fields , one tending to increase , the other tending to diminish , the total moment .
This uncertainty cnters into the application of as well as into its original determination .
For in the vibration experiment the temporary and permanent magnetisms are in the same sense , whereas in the deflection experiment they are in opposite senses .
If the presence of a permanent magnetic moment influences the alue of , then the value found at Kew , even if otherwise above criticism , presumably become less exact as the moment falls off .
Another source of uncertainty is the probability that varies the temperature .
We can ) judge of the possible influence of these sources of uncertainty by reference to the formula * Lamont , 'Handbuch des Erdmagnetismus , ' pp. 149\mdash ; 150 , is less for engthening than for weakening fields , the permanent moment is very large .
mination of Earth 's gtic connecting the error in X with the departures and of from its assumed value during the vibration and deflection experiments .
As exceeds 1 , we see that if and be equal , the first { erm on the right of the equation is the larger of the tlvo ; thus an error of given amount in has most cffect in the vibration experlment .
We have approximately land putting , we should find for average collimator magnet , at KelY approx. The average value of in Table I is ; thus if and S ' exceeded 3 per cent. of the value of the consequent error in X n-ould become sensible .
A change of C. in temperature seldom alters ) more than 2 per cent. , so in temperate regions the neglect of temperature tion in seems hardly likely to lead to error , unless of course porary magnetism is more susceptil ) to temperature changes than permanent magnetism .
In tropical countries the neglect is likely to be more serion .
[ Since this paper was written , Professor Iascart , in ' Terrestrial lIagnetism , ' has called attention to a source of uncertainty hich I had supposed negligible .
He points out that in the vibration nt the proper induction coefficient is really , where has the mea1nng attached to it here , while is the coefficient of induction for a letlC field perpendicular to the magnet 's length .
Professor Mascart describes a method of determining experimentally , and \mdash ; the coefficient when temporary and permanent longitudinal magnetisations are opposed .
The method is , he says , easily applied and the resuIts consistent ; ) the apparatus required seems a little comphcated .
Professor quotes no numerical results .
In the Kew pattern collimator magnet I should anticipate to small , but it may not be negligible .
Experimental tlOU of the point is certainly desirable .
] lIIagnets .
S39 .
In the deflection experiment it is eustonlary to describe the collimator magnet as being on the east or west side of the nurror magliet , as having its north pole east or ujest .
The directions arc not strictly cast and west , but perpendicular to the mirror nltgnet , as deflected out of the magnetic meridian .
As the sense inteIlded , , is qmte clear , I shall employ the usual terminology .
Dr. C. Chree .
Magnets and the In the accompanying figure , which is not dralvn to scale , and epresent the " " west\ldquo ; and ' ' east\ldquo ; ends of the graduated bar on Vhich slides the frame carrying the collimator magnet .
The graduations run in both directions from , the centre of the bar .
The centre cf the mirror magnet should be on the vertical line through , and also the horizontal line which is the prolongation of the magnetic axis of the collimator NS .
The figurs shows the collimator magnet in the four positions which it occupies during the deflections made at one of the two distances , say 30 cm .
Supposing a 30 cm .
deflection to be the first of the day , the four positions are assumed in the order 1 , 2 , 3 , 4 .
It is the rule at Kew 0bservatory that if a 30 cm .
observation takes precedence on a given occasion , then a 40 cm .
observation has precedence on the next .
On the second occasion the four 30 cm .
positions occur in the order 2 , 1 , 4 , 3 .
The object is very probably to eliminate so far as possible the influence of changes of force or temperature on the value of , which is calculated from a large number of observations .
S40 .
If there were perfect symmetry one have identical readings in positions 1 and 4 , and again in positions 2 and 3 .
But , in reality , agreement in these readings is wholly exceptional .
To eliminate the disturbing fluenco of observational errors and changes in force a of deflection experiments are necessary .
Thus , the only case in which I have had the data 1equisite for an exhaustive study is that of the Kew unifilar itself .
In this instance the phenomena seem adequately explained ) the hypothesis of the existence of such asymmetry as we should anticipats a priori .
It is doubtless the aim of the maker to fit the suspension tube so that the centre of the mirror magnet and the ccntre , or zero , of the graduated bar in the deflection experiment shall lie in a vertical plane perpendicular to the bar ; but complete success is mikely to ttend his attempts .
Accordingly , my first assnmption is that the centrs of the mirror magnet is really at a distance , counted positive towards the " " east from the vertical plane to the bar through the centre O. The frame carrying the magnet has a tiducial mark ' at the centre of its edge , which is set the deflection experiment at the of Earth 's ) ' ) division 30 or 40 cm .
on the .
From the celltre of the upper edgc of the frame a pillar extends vertically , so as to fit into the hollow shank which forms part of the appendages of the magnet .
The vertical plane to the ) I the fiducial mark- olltain the centre of the vertical pillar and also that of the magnet .
Supposing the to fiducial marks 011 both sides , and that these agree , one can test the istence o the 1St possible of asymmetry as mmetr turning the frame end for end , keeping the direction of the tlnc hang ; the azinlnth of the lllirrol magnet serve as a criterion .
This kind of asymmetry has jl0t been detected in the Kew unifilar ; and if existent its fluence n-onld be nearly elinlinated in a large lmber of 1n which eithel position of the sliding fralne occurred pronuscuously .
ex stcnce .
Th of the second species of is less easl 1 tested but is more probable a priori .
Acoordingly my seconld assunlption is the deflection expel'iments the centre of the magnet a ) from the vertical plalle n passes the fiducial ) , and is perpelldiculal to the magnet 's .
Here ulted p when the 's centre and n pole are 011 the side 01 the fiducial mark .
S41 .
On these two hypotheses it from fig. 1 , that the apparent ) etween the centres of the gnets is , the true distances are as :\mdash ; 2 .
3 .
4 .
Position .
1 .
True distance Now ose and are the deflection angles of the ow mirror magnet answering to the true and ) : Then , according to the first approximation ( S)3 ; whence , neglecting tres and products of ) ( 10 ) .
In what follows I employ when cm .
, and cm .
, and distinguish the several positions of fig. 1 by sllffixes .
to the scheme of values of above , find\mdash ; Positiou 1 .
Position 2 .
tall Position 3 .
Posibion 4 .
Dr. C. Mcagnets and the lVhence The mean resuIts from the thirty-nine absolute determinations of horizontal force made with the Kew unifilar in 1892 , were as follows:\mdash ; 10 ' Also , approximately , Thence we have\mdash ; From observations at 30 cm .
cm .
, whence cm .
cm .
observations at 40 cm .
cm .
, , , whence cm .
cm .
If the hypotheses on which the calculations are based are incorrect , the agreement between the two sets of values found for and is certainly remarkable .
I repeated the calculations for the data from the same unifilar in 1893 , and again the two sets of values for and were in excellent agreement , the means being cm .
, cm .
S42 .
of the first approximation we have of course when we neglect and Thus to estimate the size of the error we must go as far as squares products of small quantities , replacing ( 10 ) by ( 11 ) .
Substituting the first approximation value for in the small terms in the usual way , we find ( 12 ) .
Taking , and ascribing to in succession the values corresponding to the four positions , I find on reduction ( 13 ) .
method followed above has the advantage of leading directly to the alue Determination of the Earth 's .
407 Taking the data for the Kew unifilar in 1892 , we find ( or about 48 Now is the error , in circular measure , in the deflection and the consequent error in the horizontal force , deduced from the relation constant , ( 14 ) .
In the of the ifilar , we have approximately X whence SX C.G.S. , approx. Thus the error is nearly one unit in last significant figure .
As the error varies as , it would become of very sensible magnitude if and were twice or thrice as large as in the Kew mifilar .
If I may judge from the figures recorded of unifilars under tion at the Observatory , such largs values of and really exist .
The sourcs of error just considered alters the observed deflections at 30 and 40 cm .
in the same sense , and seems unlikely to exert an appreciable influencs on the value of P. S43 .
Causes of mmetry.\mdash ; A collimator magnet ought to be adjusted in its stirrup until it is exactly in the vibratio1t experiment .
The magnetic forces acting on it tend , in this magnetic hemisphere , to pull the north pole down , and consequently the centrc of gravity of the composite body composed of magnet and appendagcs must be out of the vertical , which is the prolongation of the suspending fibre , and on the same side of it as the south pole .
If the stirrup is symmetrical , and if the lens and scale of the magnet are equal in weight and symmetrically situated , then the C.G. of the magnet itself must , in this hemisphere , lie on the same side of the snspending fibre as the south pole .
Now the suspending fibre in the vibration experiment , if produced , should cut the magnetic axis at the same point as it is cut in the deflection experiment by the vertical plane which is perpendicular to the magnet , and the fiducial mark on the sliding frame .
Thus in a perfectly symmetrical magnet the middle point of the magnet 's length , and so the middle point of the line joining the " " poles\ldquo ; assuming ( metic symmetry ) must lie on one side of the fiducial mark , and so of the graduation on the bar which coincides with it .
Calling the error thus introduced in the distance between the of , a quantity wnose physical significacce is oasily grJsped .
Bnt a sulItCiently exa value of can be obtained more simply from the relation .
408 Dr. C. Chree .
Collimator and the magnets ' centres , measured as in S40 , we can find its value if we know the circumstances of the case .
Thus suppose the magnet to havs !
a moment 840 and to weigh , with lens and scale , 30 grams ; then at a place where ( gravity ) is 980 , and the vertical magnetic force is , we have cm .
, approx. The value thus found for is not only much less numerically , but even opposite in sign to the value calculated for the Kew collimator magnet in S41 .
There is nothing surprising in this , because the perfect symmetry which our last calculation assumes in the magnet and its appendages is a remote probability .
For instance , the lens and scale of the magnet , in the only case where I had them weighed , differed by nearly half a gram .
If other things were symmetrical this alone vould remove the C.G. nearly cm .
from the symmetrical position .
Mechanical asymmetry might arise in many other ways .
The magnet itself , for instance , might be slightly conical .
It is also at least conceivable that magnetic asymmetry may exist .
If a magnet were conical , or had one end tempered diflerently from the otlJer , see no reason to expect its " " poles\ldquo ; to be equidistant from the middle point of its length .
S44 .
Whilst numerous causes may contribute to the asymmetrical displacement denoted by , the first mentioned appears the most interesting .
Not merely is it , as we have seen , unavoidable , but it varies if either 7 or the vertical force alters .
The vertical force on the earth 's surface varies from about to C.G.S. unit .
Thus if ths typical magnet above considered were adjusted so as to bs horizontal under the one limiting force , it would have to be shifted about cm .
in its stirrup to beoome horizontal under the other .
The shifting in the stirrup entailed by this variable cause of asymmetry slightly affects the moment of inertia .
For instance if the symmetrical magnet considered above , for which cm .
at Kew , were used in a tropical station where the vertical force vanished , should be reduced to , and the consequent change in moment of inertia would be C.G.S. unit , approx. The mean in Table I is , or 274 approximately ; and with these values , approx. Supposing no allowance made for this , the consequent error in X would be Determination of the Earth 's Ilorizonted Magnetic ce .
4(I9 This would hardly be sensible supposing X measured as usual to five signfficant figures .
It might happen however that owing to asymmetry in the stirrup the C.G. of the magnet\mdash ; with scale and lens\mdash ; had to lie , in the absence of vertical force , at a distance , comparable with 1 mm. on one side of the suspension fibre .
Under such circumstances the change in the moment of inertia when the vertical force altered from to would exceed that calculated above in the ratio As perfect mechanical symmetry must be the exception , we see that this source of uncertainty is hkely to be considerably more serious than might be supposed from our first example .
The above considerations may suggest that shifting the collimator magnst in its stirrup is a mistake .
If , however , this is not done when required to keep the magnet horizontal , the magnet and appendages must tilt over slightly .
This not only alters the moment of inertia , but likewise prevents the magnetic couple in the vibration experiment from having its proper value .
The source of trouble we have just been oonsidering is distinctly exceptional in this respect , that it is most serious in mag1lets of large magnetic moment .
Law of Action betue S45 .
The formula couple assumes that the centre of the mirror magnet lies on the prolongation of the axis of the collimator magnet , and that the axes of the two magnets are perpendicular to one another .
Vith the magnets in the position supposed , it is not difficult to calculate the values of and we can regard a magnet as consisting of two equal and ' ' poles the product of whose distance into the strength of either constitutes the magnetic moment .
Thus if and be the distances between the poles of the collimator and mirror magnets respectively , I find couple , so that ' ( 16 ) .
lVhen cm .
( 17 ) .
Dr. C. Chree .
Collimator and tlve We have when , , , approx. can thus make either or vanish , but not both simultaneously .
YVhen S46 .
Both collimator and mirror magnets are comparatively " " short , their lengths are only about ten times their diameters .
In such magnets we should expect , according to Coulomb , where and lt are lengths of the magnets .
The formulae by Green , Jamin , and others are more complicated .
Perhaps all we are justified in saying , a priori , is that and cannot exceed and l'/ 2 , and that the collimator and mirror magnets are sufficiently similar in pattern to make it probable that the assumption is not far wrong .
To throw some light on the question , I had measurements made of the magnets of the Kew unifilar and of a unffilar of class E. The Kew lmifilar belongs to group , and its is negative but exceptionally small .
The instruments of group are , as we have already seen , exceptionally uniform in character , and invariably have their P Iarge positive .
The results were as follows , lcngths being in centimetres , Unifilar .
' .
Kew , group 9 .
6 .
0.69 Supposing , we should have in these two cases 0 .
0 .
0 .
These figures , taken in conjunotion with our previous data , are on the whole distinctly favourable to the hypothesis ( a constant ) ( 18 ) Accepting this hypothesis provisonally , I had the curiosity to see what value we should obtain for by ascribing to the mean value found for unifilars of group , tnking for and the values quoted above for a of Determination of the Force .
411 Equating to one another the two values of at 30 cm .
, the one as given above , the other as given in Table I , we I we Thus , approx. , and so , approx. The practically exact agreement with Coulomb 's relation seems to warrant the conclusion that if we employ this relation in calculating the result is likely to be of at least the right order of magnitude .
On this hypothesis , taking the values and for fairly representative of the two groups , we have with cm .
, Probable value of in unifilar of group , approx. \ldquo ; , , Assuming the value of to be correctly dc.ermined , the error in X due to the omission of the term is given by Taking the numerical values found ) for , we should have when For mean unifilar of group , SX , approx. , SX , , There is admittedly much uncertainty in these numerical estimates , but they undoubtedly indicate that the neglect of requires justification .
If the term is not negligible then the ordinary method , which assumes that the term is the sole cause of the difference between the two values of , given by the first approximation formula and , must lead to a wrong for P. Now the correction does not strike a mean between the resuIts at 30 and 40 cm .
, but adds to or subtracts from both values of Thus the indirect consequences of neglecting may be as important as the direct .
[ Lamont , in his ' Handbuch des Erdmagnetismus ' ( Berlin 1849 ) , discusses the general case of the action of one magnet on another , the distribution of ' ' free\ldquo ; magnetism being .
The results ( 16 ) for formulae , the relative importance of , and the values of , for which and vanish , must vary according to the law of distribution of free .
At the same time , it would appear that so long as the law of ltion is the sams for the two magnets , the conditions for vanishing is hardly likely to depart much from The idea adopted in the text of assuming " " poles and proceeding on the hypothesis constant , is one of which Lamont says on his p. 45 , " " Dies ist ein sshr nutzlicher Mittel um approximativ den Werth der oheren Glieder zu bestimmen . . .
Thus , though not new , as I had supposed , it has the advantage of being recommended by one of the leading pioneers of terrestrial magnetism .
] S47 .
The exact symmetry in the position of the two magnets which is postulated in obtaining the formula for the deflection is rather to be hoped for than expected .
For instance , the magnetic centre of the mirror magnet is hardly likely to lie exactly on the prolongation of the magnetic axis of the collimator magnet , nor is the magnetic axis of the mirror magnet likely to be exactly normal to the mirror .
I have , however , no sxperimental data bearing on either of these points .
S48 .
The present paper may seem premature , in view of the number of questions which it raises without finally answering .
To wait , however , until experiment had given a final answer to all the questions would simply mean shelving the whole subject indefinitsly .
Many of the questions could be satisfactorily dealt with only by elaborate expsriments .
Such experiments are hardly likely to be carried out at any public institution , within a reasonable time , until the necessity for them has been clearly demonstrated .
The investigations embodied in this paper have extended over several years , and the results obtained are likely , I think , to be of immediate use .
The unifilar magnetometer is not the only instrument capable of measuring the horizontal force .
Induction coil methods have of late years been introduoed and advocated by eminent authorities .
The question of the relative merits of magnetometers and induction coils is very likely , I think , to come to ths front in a few years , and it is hardly Absorption of Rontgen 's Rays by Solutiom of Metallic Salts .
413 likely to be adequately considered unless the sources of weakness in the respective methods are clearly understood .
Again , there has been an increasing tendency of late years , in most lines of physical research , to add figure to figure , and to aim at higher accuracy , or at least the appearance of higher accuracy .
If people were content with giving magnetic force to four significant figures , the majority of the sources of uncertainty specified in this paper need awaken little apprehension .
But now-a days hardly any one is content with less than five significant figures , and one occasionally sees six .
I am not prepared to say that the retention of six figures is indefensible in the case of open range magnetographs , when it is clearly understood that rences only are concerned , but I do think that with unifilars of the Kew pattern , as hitherto made , the least )robable error we can reasonably expect in an absolute measurement two or three units in the fifth significant figure .
S49 .
We have seen that there is every reason to expect that the values of the horizontaI force given by a unifilar are in error to a different extent , according to the temperature and the magnitude of the force .
must introduoe a source of uncertainty into the comparisons effected between unifilnrs at distant observatories through the of a travelling instrument .
To adequately forewarn those engaged in such comparisons may be to put them in a position to obtain more satisfactory results .
S50 .
the opinions expressed in this paper , and for the accuracy of its conclusions , I am alone responsible ; but it is only proper that I should acknowledge the valuable assistance given me by Mr. T. W. Baker , Chief Assistant at the Observatory .
For many years .
Baker has taken the great majority of magnetic force observations at Kew , and his experience in determining the " " \ldquo ; of collimator magnets is probably quite unique . .
Baker has always been ready to place his great practical knowledge of the subject freely at my disposal , and I have also to thank him for carefully carrying out a variety of special experiments , made to elucidate doubtful points at various stages of the inquiry .
" " The Absorption of Bontgen 's Puays by of Metallic Salts By the Right able L BLYTHSWOOD , .D .
, and .
W. LIAltcHANT , D.Sc .
Communicated by KELVIN , F.R.S. eceived March June 15 , 1899 .
The absorption of X rays by metallic salts is a bject that has not Teceived very much attention up to the present time , although it appsars to be of considerable importance .
It secmed possible that if ths

