]\gt ; On Periodicity in Sunspots .
By Sir JOSEPH , Cambridge , and N. YAMA , University of Tokio .
( Received March 28 , 1917 .
) The most important index to the dynamical constitution of the Sun lies in its dark spots , which have been known , ever since their by Galileo , to recur periodically .
Numerical data for the numbers of spots simultaneously present , and later and more precise data for the extent of spotted area , are in existence for about two centuries ; and it is natural that should be made , from all points of view , to extract the knowledge which they contain .
The discussion of the fundamental question , whether there is permanent unbroken periodicity , due either to planetary influence or ( as seems much more probable ) to a period of ciynamical oscillation , to the Sun itself , was taken up statistically by Dr. Simon Newcomb in a paper " " On the Period of the Solar Spots * His criterion was to examine whether the more precise phases were equidistant throughout the record , or on the other hand their deviations increased continually according to the law of errors .
The phases chosen for scrutiny were four , those of maxima and minima and the two more definite intermediate times of mean spottedness .
An analysis by the method of least squares led him to an unbroken period , of years , and thus of great definiteness .
His conclusion is that " " underlying the periodic variations of spot activity there is a uniform cycle unchan , from time to time and determining the general mean of the activity But to get this very remarkable degree of precision he had to reject the records belonging to the two decades around the year 1780 , which showed violent ularity in the phases .
" " I was at first disposed to think that these perturbations of the period might be real , but on more mature consideration I think they are to be regarded as errors arising from imperfection of the record .
The derivation of any exact epoch requires a fairly continuous series of derivations made on a uniform plan .
If we compare and combine the results of observations made in an irregular or sporadic way , ib may well be that the actual changes are masked by the apparent changes due only to these imperfections And , again , " " it would seem from what precedes that a revision of the conclusions to be drawn from the observations of sunspots during the interval of 1775-1790 is very desirable In the raphical anaiysis which foilows , we shall be driven to reject the 'Astrophysical Joul'nal , ' vol. 13 , pp. 1-14 ( 1901 ) .
Sir J. Larmor and Mr. N. Yamaga .
record of the two cycles of the spots , here referred to , for much more conspicuous and emphatic reasons ; so violent a temporary interruption as they show , in the course of a phenomenon periodic in phase and on so vast a scale , seems to be most anomalous .
It is true that the graph ( fig. 1 ) shows shrinkage in numbers of spots recorded in not a few cycles , involving change of ordinate smaller indeed but yet of the same order of magnitude as the discrepancy in those two cycles ; but in no other case is the period deranged .
All the rest of the graph is consistent with a law represented by the product , where is a function strictly periodic in years , and is a slowly changing factor ( cf. fig. 9 ) which determines the varying amplitude .
The next analytical discussion is a memoir by Prof. Schuster , " " On the Periodicity of Sunspots , which applied the method of quasi-optical analysis elaborated by , in order to determine the periodogram of the disturbance , , the curve connecting amplitude with frequency in the continuous of simple undulations into which the data for a specified range of time , taken by themselves , can be analysed .
memoir is the most complete and important analysis that has yet been made , and the data marshalled in it from Wolf and Wolfer , and from the records of Greenwich and of the Solar Physics Observatory , have been largely employed in the present discussion , The very striking , even at first sight startling , result is reached by him , that when the data for the last available 150 years are divided into two equal groups , their rams come out totally diiferent ; the more roup showing a nearly pure simple undulation of Newcomb 's period years , while the earlier roup hardly shows this period at all , but exhibits prominently two somewhat broad and indefinite periods of about and years .
Yet Prof. Schuster also assigns much to Newcomb 's feature , that the turning points and mean points on the raph of spot-frequency occur at equidistant times throughout the , except for the one conspicuous but transient local disturbance after which the phases recover the previous sequence ; he considers that this is strong evidence that the -year period thence predicted is a permanent property over the whole range , and of some sort of dynamical type .
One of the chief points elucidated in the present more elementary and direct discussion is the cause of the disappearance of this period of years on the analysis of the record from 1750 to lS20 .
It is that the two spot-cycles , already referred to , between 1776 and 1798 , which are conspicuous and therefore greatly influence the analysis , are entirely abnormal .
If this range of the data is left out , the rest of the 'Phil .
Trans February , 1906 .
On Periodicity in Sunspots .
record falls in satisfactorily with a permanent -year period .
This is brought out especially by the graphs which show , erimposed , the forms of the periodic components this peried , that have ) extracted from the three ranges separately into which the record has ) divided for the purpose of this test .
More recently , Prof. Kimura*has obtained , ) a process of adaptation , series of trains of unlinlited simple undulations which closely represent the sunspot graph over the 164 years of available record , represented here in fig. 1 .
A free curve drawn through the ordinates representing these discrete amplitudes , at equi-distant frequencies , ] trend vards a .
In two following papers , Prof. H. H. Turner has pointed out that this group of components coincides closely with the ]ourier series representing this record of the 164 years , repeated , however , without end .
This coincidence is not a physically necessary one , for the components determined from a longer range would , of , include this range , not however repeated in the same manner .
And the curve will not be the same either , unless the material is homogeneous , in a statistical sense of which such identity would be the test .
Prof. Turner points out that the presence of a periodicity sustained over a considerable length of the data , but not identical with one of these Fourier harmonics , would tend to be revealed by abrupt of in the range of amplitudes ; and he shows how the vnlue of this period might be deduced , provided .
however , it belongs to an unending simple train .
The number of spots is necessarily positive ; but , as it fades to zero -ell as rises to a maximum , it may be held to be a measure of the energy of the solar disturbance , which is proportional to the square of an amplitude purely alternating from positive to ative , in the simplest case say to which is cos2 pt , representing oscillation around a mean .
At first glance , this view may not be quite in with the antisymmetrical character of the periodic part of the graph , which rather indicates a mean spottedness with this special type of periodic fluctuation superimposed on it ; this other special feature , that the total spottedness falls sensibly to zero at the minimum before again , must also be kept in view .
The very fundamental character of the solar phenomena that must be concerned in the spots will justify the application of a direct method , involving perhaps fresh features and principles , to this well-worn problem .
A considerable selection from our graphical representations has been here 'M .
N. Roy .
Astron. Soc May , 1913 .
Sir J. Larmor and Mr. N. Yamaga .
included , as the inferences rest more on visual inspection than on calculation and , moreover , experience has shown us that it will be convenient inquirers who in future approach the subject , to have ready at hand the graphs which constitute the material of the problem .
The iven in fig. 1 represents the sunspot record , year by year , from 1750 to 1914 , the data for the last years from the Greenwich record having now been added to the numbers as tabulated by Dr. Wolfer and the as recorded by the Solar Physics Committee and at Greenwich up to 1902 which are the data that have been discussed by Newcomb and by Schuster ordinate in the raph is the mean of three consecutive monthly records so as to give a mean record all round the Sun instead of mainly on one henlsphere .
The broken arrows mark off the Wolfer-Newcomb-Schuste ] cycles of years ; the full arrows represent an altered trial period years , which perhaps fits more obviously with the phases of the mosl recent data .
In any case it will be of utility to have analyses for two tria1 periods .
The misfit of both sets of arrows between 1776 and 1798 is obvious A smooth mean curve has been drawn through the peaks of the numerical data in The discussions of Wolfer , and Schuster having afforded strong evidence for a very definite secular periodicity unbroken as regards phase its amplitude in the sunspot graph still remains for consideration The procedure employed here is to multiply the amplitudes throughout each cycle by a common factor , so that all the cycles shall become equal in mean amplitude .
The result is represented in fig. 2 ; the multipliers required tc deduce from fig. 1 are represented , smoothed out into a curve , in fig. 9 , in which each short line represents the actual constant multiplier for a FIG. 9 .
Curve amplitude cycle .
If represents , and represents fig. 1 , , which is the actual sunspot graph , then .
The curve of in fig. 9 shows that this amplitude factor gradually throughout the whole range of the data .
Thus , if we assume that On Permanent Periodicity in Sumspots .
the -year constituent of the sunspot record is of constant period but varying amplitude , it will be represented with fair fidelity by , where is the periodic constituent of this period that is inlvolved in the modified graph of fig. The method of extracting a secular constituent of umbroken period from graph is cruratively ) to wrap the graph round a cylinder whose cilcumference is , and take the mean ordinates of the various convolutions to form a new curve .
All elements of an graph which have not a period at , or neal ' it , will then out ; if the range of the graph that is utilised is a large multiple of , the result may be presumed to represent with fidelity the periodic constituent required .
This periodic part , of form thus isolated , may then be analysed into a Fourier series representing the fundamental simple period and its submultiples .
It will readily appear , on consideration of .
the nature of this the values of these simple harmonic periods of equidistant frequency .
thus determined , will agree with the ordinates in a continuous periodogram*which represents the actual disturbance over the long that is thus analysed , taken , however , by itself , with nothing added outside that range .
But we are now dealing , by hypothesis , with the different case of a definite period repeating itself outside that range as well as inside ; its components will represent unbroken periods , and the periodogram for them will be , strictly , a series of isolated narrow ) eaks with vanishing intermediate amplitudes ; if the rule is retained that the amplitude of a compon8nt of range is the element of area of the , these isolated ordinates would have to be infinite in .
The continuous periodogram curve thus cannot quite naturally represent sharp permauent periodicities , and a different procedure as here , in fact a much simpler one , is as regards them to be preferred .
On the other hand , the periodogram analysis , as elaborated by Prof. Schuster , is well adapted to the discussion of material containing transient periodicities repeated irregularly , but with the same periods , or tendencies to periods , recurring .
The standard example is natural radiation .
If two reaches of the material that is analysed , of sufficient lengths , lead to like rams , then to that degree the material may be presumed to be statistically homogeneous ; otherwise it is fortuitous and sporadic .
It seems to be inferable from Prof. Schuster 's analysis that the sunspot record , apart from the year * With amplitude plotted against frequency , not length of wave .
Sir J. Larmor and Mr. N. Yamaga .
period , is in the latter Yet it is at first sight startling that the -year period is almost non-existent in the periodogram of the first of the two 75-year ; for , although that range contains two anomalous cycles of 22 years , the other cycles are normal , and might be expected to show in the result .
But such destructive interference is really not anomalous .
For example , if a range of length contain a train A , limited to length which is a considerable number of periods , this train will , irrespective of its position along the range , add to the periodogram at its own period a term approximately , exhibiting there a maximum ordinate of phase .
If the range contain a number of such limited trains of uniform amplitude , their contributions can be summed by the vectorial raphical process into a single term If there are many of them , and their phases are sporadically distributed , the resultant of their vector amplitudes will be very small compared with the arithmetic sum , and will often be quite insensible .
But if we include a succession of very many such ranges , and the phases of the component trains vary sporadically all along , then , by Lord Rayleigh 's principle , the total resultant alnplitude has a sharply defined probable value , equal to the most probable value for a , multiplied by the square root of the number of , a value , in fact , determined from the total energy along the range , treated as additive .
The optical analysis of radiation is an , then , rendered definite by ranging over an extremely great number of periods ; we can , in general , expect nothing so definite , unless under exceptional circumstances , for shorter ranges of graphs , such as are alone in observational material or amenable to Fourier analysis .
The constituent of period years , as determined from the monthly sunspot numbers raphed in fig. 1 , by procedure of superposition of periodic lengths , conducted arithmetically ( but modified by omission as infra ) , is represented in fig. ) and .
The process has been arranged so as to proyide a test for the hypothesis of unbroken periodicity , by dividing the range of the data into three nearly equal parts , extracting the graph of the periodic constituent from each of them , and taking the mean of these three graphs superimposed in fig. 3 , duly weighted if necessary , for the final result in .
The more nearly concordant these raphs in fig. 3 turn out to be , the more justified is this hypothesis as against sporadic broken periods .
The hest trial shows , however , at once that this hypothesis , otherwise quite satisfactory , is wholly upset in the two abnormal and 1 * On this general subject , see Sir J. Larmor " " On the Fourier Harmonic Analysis in Presidential Address , ' Proc. Lond. Math. Soc pp. 8-42 , November 2 , 1916 . .
cit. , ' Proc. Lond. Math. Soc p. 35 , November , 1916 .
FIG.4 .
Final mean curves : and doubled depa r Frorn FIG.3 .
Periodic components , ried data . .
Soc. Proc. , A. vot .
93 .
1o so 30 40 50 60 70 80 90 .
Fi nal mean cu rve : and doubled from an On Perioch city in Sunspots .
cycles shown in between 1776 and 1798 .
The course very strongly suggested is to omit them from the data to be analysed , as for some reason wholly foreign to the problem ; this will in no way affect the validity of the analyses of the rest of the sunspot graph , but the question will remain over , whether there has been error of record or whether some overmastering transient cause has completely masked the normal run of the data .
When the magnitude and the time-scale of the solar phenomena that are involved is borne in mind , the former alternative seems to be the preferable one ; it was the one adopted by Newcomb for reasons quoted The two sunspot cycles from to 1798 have been omitted and the remainder of the data divided into three groups of nearly equal duration : to 1822 , to \mdash ; 1822 to 1867 , 1867 to 1912 .
The constituents of period years extracted from these groups of the original data ( fig. 1 ) are recorded in fig. 3 , that from the first group being represented by the broken line , that from the second by the continuous line , and that from the third by the dotted line , as indicated above side the ranges of time .
It will be seen that these three , raphs do fairly superpose ; fact that this coincidence is not very good , even when allowance is made for the , small number of cycles in each .
group , arises In reply to inquiry as to whethel the records could possibly be so greatly in error for these two cycles , Prof. Wolfer kindly referred us to his classical memoir in the Zurich ' chrift , ' 1902 , where in pp. 80 seq. he discusses the discrepancies between the various early records ; nothing , however , so substantial as is here in question seems to emerge .
At the end , p. 95 , Prof. Wolfer recalculates the 11-year period and its och , a result ( for the period ) in close agreement with Newcomb 's then recent one , notwithstanding variations in detail of the data employed , and substantially in agreement with earlier determinations Wolf and Sporer , and by himself in 1892 .
Prof. Wolfer has also constructed a graphical representation , like fig. 1 here , on a very open scale , the whole range ( : Zurich Astr .
Mitteil vol. 93 ; 'Monthly Weather Review , ' April , 1902 ) , which does not any more than fig. 1 give ground for the suggestion of faulty records , atever may be said as to the minor matter , from the poiut of view of our present discussion , of indetermination of the principal phases .
Refel.ence should also be ntade to the ' Meteorolog .
Zeitschrift , ' vol. 5 , 1915 , where Prof. Wolfer arrauges the sunspot data for the interval 1901-1914 .
Earlier knowledge of these references would have facilitated and probably proved our present : but the main issues rise in it are on a large scale , and would not be sensibly affected by change in the details of presentation .
[ See addition , p. 503 , infrct .
] steep rise and gradual fall is a well known feature of each cycle separately .
Sir J. Larmor and Mr. N. Yamaga .
mainly from the presence of low amplitudes in a range of the cycles ( 1800-1834 ) mostly in the first .
The unweig mean of the curves for the three roups , which epresents the whole of the data , is recorded in fig. 4 , both as a broken graph representing the actual numerical record , and as a smoothed curve such as a physical periodic constituent would naturally be .
It will be observed that the smooth curve is here a good representation of the broken graph ; and this is in part because the cycles of low amplitudes are now swamped by the greater number of cycles with amplitudes more normal .
We have also extracted , for comparison , the periodic part from fig. 2 in which the amplitudes of the cycles have been equalised .
The results are shown in and .
The three constituent ranges in fig. 5 now agree much more closely ; but their mean , fig. 6 , differs little from the previous form , fig. 4 , for a reason already assigned .
In the two latter curves ( fig. 5 and fig. 6 ) , the period has been taken as years ( somewhat at random , of .
p. 496 ) instead of , as was to be anticipated ( p. 501 infra ) ; this slight of period does not sensibly affect the form of the periodic constituent , and it is unnecessary to exhibit graphs for both periods .
The Fourier analysis of the periodic part of the unmodified sunspot curve thus extracted , is , in arbitrary units , and of the same curve modified to equal amplitude is where the periods chosen are , or years , and , or years .
Terms in the series with amplitude less than unity are omitted , as being within the limits of uncertainty .
With this limitation , no cosines appear in the expressions .
Thus , when the origin is taken on the periodic curve at Dliddle , that curve is of type satisfying the relation , the positive side of it the reversed image of the negative side .
As this antisymmetrical character is remarkable , and probably estive for physical theory , we have confirmed it by constructing a graph of each of these Fourier series and fitting these graphs on to the original curves from which rhey were derived .
Nowhere is the discordance more than onethirteenth of the maximum ordinate , and that amount is very local ; the curves are in the main so close that the ifference would hardly show on diagrams of the size here given .
The antisymmetrical character of the periodic graph that is thus revealed ' may also be scrutinised directly , without the analysis into Fourier components .
Any graph whatever can be divided into a symmetrical part and an On Permanent Periodicity in Sunspots .
antisymmetrical part with respect to any , these even and odd functions being determined by the equations .
In the present case the symmetrical component , equal to , is at any rate small .
Its value can be plotted by folding the periodic graph over on itself , and then the mean ( half sum ) of the two ordinates , one positive , one negative .
The results thus obtained for the rugged } ) representing directly the numerical data , and for the smoothed curve , are iven by the two flat curves drawn above the periodic graphs in figs. 4 and 6 , when the ordinates in them are reduced by halving .
These curves , thus halved , by comparison of their ordinates with the that were smoothed out in the periodic raph , provide a criterion of the degree of certainty of the property now under discussion .
If it is regarded as confirmed , a slightly purified value of the periodic raph itself would be given by .
If figs. 3 , 5 , and 4 , 6 are supposed shifted along into contact , each pair will represent two of the period , and the lower bend of the curve may be seen by eye to be of the same form size as the image of the upper bend .
We have yet to show that the process of reduplication of the raph , in order to extract permanent periodic components , can be relied on , i.e. , that the effect , on the periodic raph so obtained , of a small error in the true period is igible .
If the true periodic part of the graph , so that , the use of a trial period will produce from the material instead of , a quantity substantially equal to divided by the number of or cycles in the data .
As is small this is approximately The ordinate as computed on the trial period is thus the mean value of the ordinate of the true periodic part over the , on the positive side of when is positive .
So long therefore as is small compared with , an error in the period will lead to but slight error in the form of the periodic component thence derived ; for the presumption is that the true periodic part is represented by a smooth curve without violent bends .
An illustration is afforded by the result of the trial periods and years and fig. 6 .
Had their data been the same , instead of being and .
the criterion of direction of the error would have indicated that the latter may be Sir J. Larmor and Mr. N. Yamaga .
earer the correct period .
For the amplitude derived from a trial period will be nlaximum when it coincides with the true period .
The question arises whether the constituent of period years thus determined is the only continuously periodic element in the sunspot record .
Other likely periods might be tested by the same process , but there is the disadvantage that periods commensurable ( or nearly so ) have a part in common , made up of their common harmonics , *which may have to be separated out .
It is more convenient and suggestive to take out from the record the periodic part already found and analyse the residue .
The residue for the natural graph ( fig. 1 ) is given in fig. 7 , and that for the modified raph of equal amplitudes ( fig. 2 ) is iven in .
In these residues , two defaulting cycles are , of course , conspicuous .
The residue when fundamental simple constituent alone of the -year periodic part has beer removed , is given as a raph in Prof. Schuster 's memoir ( p. 92 ) , and peaks that it reveals ( mainly , however , in its first half , which contains two abnormal cycles , \mdash ; they would hardly be noticed in the second half itself ) rgest to him a possible permanent period of years .
Ou ] lesidue , when the whole of the year period is removed , has been tsste for this period by the same process as above , dividing up into three groups The results are shown in : that of the third group containing the mosl FIG. .
Components Fo year .
modern and exact data undulates closely along the zero axis , and thus shows no such period ; the first and second roups both show rises from zero In the optical analogy of spectroscopic analysis by a grating , this corresponds to the overlapping of spectra of different orders .
On , in Sunspots .
axis at the beginning and end , and thus have something in common ; but when the two abnormal cycles are omitted , the first roup gives a graph close the axis , as did the third ; the second result , thus remaining unsupported , is from its form most probably accidental .
We thus find no indication of a permanent period near years .
Our conclusion is that a secular unbroken sollr periodicity of about years , controlling superlicial solar phenomena , is established firmly , if we may omit from the record of 160 years the two discrepant cycles from 1776 to 1798 .
We have found that the graph of this periodic constitueut has the remarkable property of being antisymmetrical with respect to the axis which bisects its undulations , each undulation equal areas above and below and the curye repeating as an inverted image ; which may contain a clue for a physical theory .
We hay found no indication that the residue of the sunspot activity , after this periodic palt is .
is of other than fortuitous sporadic character , not amenable to further analysis as a whole .
This latter result is supported by the totally different characters of the rams found Prof. Schuster for two successive ranges of 75 years , which favour the extended conclusion that the residue involves no sporadically repeated transient periodicities such as arise from recurrent damped oscillations , as well as no unbroken secular periods .
[ Added March 29.\mdash ; We are permitted to append a letter from Dr. Wolfer , giving very important information as to the value of the data on which discussions like the present one depend .
The uncertainty which he emphasises as to the units of measurement along the of time , seems fully to justify our omission of the very widely discrepant interval , and also enhances the interest of the analysis of the data of fig. 2 , in which the amplitudes of all the cycles have been artificially equalised .
The custo1nary mode of discussion , which includes epochs or phases alone , eliminates this uncertainty arising from varying scale ; though , of course , individual epochs are liable to non-cumulative displacement by the residue of disturbance , especially when the cyclic amplitude is small .
The results of a complete revised discussion of all available data , such as Dr. Wolfer promises , will be eagerly awaited .
" " enossische Sternwarte , Zurich , March 19 , 1917 .
" " In reply to your letter of March 14 , I take pleasure in giving you some particulars about the sunspot statistics for the critical interval from 1775 to 1795 , as far as they are at hand .
This interval has always been and will remain a very unsatisfactory one , the spot observations during that time being very scanty , and depending almost exclusively on one single observer , VOL. XCIII.\mdash ; A. 2 Sir J. Larmor and Mr. N. Yamaga .
Staudacher , of Nuremberg .
On the other hand , just for that reason , they are more eneous than many others of rermote times , and fortunately th are somewhat more numerous at the time of maximum solar activity than for the minimum epochs , which seemed to be of little interest to the observer .
Therefore , the epochs of spot maxinla ( and 1787 ) as determined by Wolf resting upon a tolerably good base , and their uncertainty does not exceed , probably does not even reach , a year .
The large departures of the lengths of the periods observed between 1770 and 1805 from the mean value of years are certainly to be considered as real , and could not be explained by the deficiency of the observations .
It not quite the same with regard to the amplitudes , or the relative heights of the different maxinr of that time , especially those of and .
The numbers of spot frequency are , of course , much more influenced by the gaps in the observations , and it is impossible to get reliable monthly and yearly means from perhaps 20 or 30 observations scattered irregularly over the year .
Another difficulty in the determination of the factors by which the spot numbers of different observers with different instruments arc reduced to the same unit , to make them comparable .
This requires a continuous system of observations of the successive observers , with Wolf\mdash ; as normal erver and as unit\mdash ; and extending backward as far as possible from later observers to former ones .
Wolf , in his investigation , did not employ all the available corresponding aCions for the evaluation of these factors , and it is probable that the factors are , in some cases , uncertain , and require corrections .
For Staudacher , for instance , I consider it too large , and also for some other contemporary observers whose factors had to be deduced from that of Staudacher , like Oriani , on whose observations the maximum of 1778 is mainly depending .
I believe the tionally h spot numbers of this maximum are partly due to this just mentioned .
It will be felt in any ation in spot periodicity founded upon the whole complex of spot numbers\mdash ; the " " relative numbers\ldquo ; \mdash ; instead of the epochs of maximum and minimum alone .
" " A new reduction of the old obseryations , back to 1749 , based on a new determination of the reduction factors , has been begun here several years ago , but it is a big task , and has not yet far advanced [ Added April 30.\mdash ; The of the sporadic , as employed bove , may be more closely defined .
The degree in which the quality described by that term is present varies versely as the on the graph within which there is some amoullt of correlation between the underlying causes ; when a is highly sporadic , each small the curve is an On iodicity in nspots .
independent entity , practically in no relation to the adjacent except that of statistical averages .
In attempt at harmonic analysis of such a raph we must therefore expand each lenlent of the Taph into its own Fourier combine the results for the elements as the laws of chance direct .
, the any duly aph is iven by where If represents a physically correlated constituent element of the yraph , situated at the origin and of sensible value only very near the origin , of total area , the formulae approximate to Thus the local element of the graph , situated at the origin of and of area , is represented in its harmonic analysis by the integral All the ordinates in its periodogram for frequencies tend to equality , except for that are not compared with the length of the element .
This remark develops into the conclusion of and of Schuster that the periodogram of uncorrelated disturbance , or of any kind of disturbance as regards gths much beyond its range of correlation , would be a straight line , all periods being equally present .
But such a conclusion implies that the area A of the local constituent element of the graph is not null .
If that were so , and the disturbance were thus as much ative as positive , it would be representable by the , radiant of some function which vanishes beyond the element , say by ; its Fourier would thus be obtained from the previous expression by differentiation , in the form The ordinate of the periodogram of amplitudes of an disturbance would thus be proportional to the frequency , as ards components whose wave-length is .
compared with the length of the elementary graph 0 with the range over which correlation holds in the disturbance .
This agrees with the facts for natural radiation , which dynamically must Prof. J. W. Nicholson .
constituted of such alternating pulses ; the at which this law intensity varies as begins to sert itself affords a measure of the on a ray over which corr -ation extends , viz. of the mean effective length of the constituent pulses in ordinary light .
Scrutiny of the variation of the intensity of Bubens ' quartz Reststrahlen ( 1901 ) with temperature seems to indicate correlation extending sensibly over a range of about six wave-lengths of yellow light , for the case of these waves at ordinary temperatures ; but the formula of Planck with modern values of the constants , not specially adjusted for long waves , gives as much as times this range for natural radiation .
It would , of course , be possible for a graph , like the residual graph of sunspots , to be devoid of correlation over the short range that is available for scrutiny , though it would come in on a much coarser scale over a long range .
] The Lateral of Bars of Section .
By J. W. NICHOLSON , M.A. , D.Sc .
, F.R.S. , Professor of Mathematics in the University of London .
Received May 11 , 1917 .
) The present investiyaGion , o strictly mathematical in character , arose in connection with a suggestion , put forward by Prof. A. Dendy and the present author in another paper communicated to the Society , *that the siliceous deposits found on certain spicules occurred at nodes of the spicules , regarded as vibrating rods .
Ihese vibrations , being set up and maintained by the impact of currents of water on the spicules , are necessarily of the lateral type .
For the detailed examination of such a suggestion , it is necessary to obtain a comprehensive account of the positions of the fundamental nodes on a free-free bar , as dependent on the law of variation of its cross-section .
The present paper contains , in fact , the formal analysis whose results were quoted without proof in the other This analysis is of considerable generality , as will appear , and the particular examples selected for purposes of illustration , together with the manner in which the variable cross-section is dealt with , have been determined by the reqnirements of the biological application already mentioned .
One general problem is in viqw throughout the work , and it may be stated as follows:\mdash ; * ' Boy .
Soc. Proc , vol. 89 , p. 573 ( 1917 ) .

