141 On Sun-spot Periodicities.\#151 ; Preliminary Notice .
By Arthur Schuster , F.R.S. ( Received November 23 , \#151 ; Read December 7 , 1905 .
) In cases where it is necessary to separate true periodic changes from other variations , which during short periods of time often simulate periodicities , the method of the periodogram is at present the only one which can give definite results .
In view of the importance of the questions connected with the changes in the frequency of the sun-spots , I have , therefore , undertaken the considerable labour of forming a complete periodogram of sun-spot variability as far as the data at my disposal allow me to do so .
The following is a brief abstract of the results ; the investigation will be presented shortly:\#151 ; The periodogram , as already explained , is the diagram representing the intensity of periodic variations as determined from the sum of the squares of the two Fourier coefficients belonging to each assumed period .
This diagram represents for any regular or irregular change exactly what the energy diagram gives us for a luminous disturbance which is analysed by a spectroscope .
Periods of not more than three or four years ' duration may be satisfactorily studied by means of the tables of sun-spot areas which are at our disposal .
For periodic times which exceed four or five rotations of the sun , I have used the mean daily areas in each synodic rotation of the sun as collected and published by the Solar Physics Committee .
The data reach back as far as 1832 .
For shorter periods we have to go back to the Greenwich measurements , which are published for each day since 1883 .
In the investigations of periods longer than three years , I have used Wolf and Wolfer 's sun-spot numbers , which are published for each month beginning with the year 1749 , and based on a combination of spot and group counting .
The periodogram of the whole interval of 150 years showed , as was expected , a marked peak corresponding to the 11-year period , but it presented some features which rendered it desirable to investigate separately each of the two intervals of 75 years into which the whole range of time may be divided .
The result was surprising , for the two separate periodograms seem at first sight to have nothing in common .
From 1750 to about 1825 the observations do not indicate any marked periodicity of approximately 11 years , Prof. A. Schuster .
[ Nov. 23 , but two periods of 13f and 9\#163 ; years respectively .
Since that time , on the contrary , the variations are well represented by an almost homogeneous period of 11'12 years .
The apparent absence of the 11-year period in the last half of the eighteenth century cannot , in my opinion , be attributed to the unreliability of the observational data .
The method employed is specially suited to investigate possible effects of planetary configurations , which have often been supposed to exist .
I have therefore specially investigated the rotation period of Jupiter and the synodic periods of revolution of the planets Venus and Mercury .
My results do not show any trace of a planetary influence .
The most persistent period which I have found is one of 4*81 years ' duration .
It appears , with good agreement of phase , in Wolf 's sun-spot numbers both before and after 1825 .
It also shows well in the direct measurements of sun-spot areas .
The amplitude of this period seems to be about one-sixth of that of the main 11 years ' periodicity .
An increased intensity is observed for periodic times of 5*625 , 3*78 , and 2*69 years , the importance of the periods being in descending order Multiplying these numbers by 2 , 3 , 4 respectively we obtain 10*76 , 11*34 , 11*25 .
These periodicities are , therefore , in all probability only sub-periods of the 11 years ' variation .
The only remaining periodicity which is clearly indicated by the periodo-gram is one of 4*38 years ' duration , but this does not occur in the observations previous to 1750 , and only shows feebly in Wolf 's series of figures as drawn from the observations of the last seventy years .
Its real existence is , therefore , doubtful , though some prominent outbreaks during the last fifteen years may be associated with it .
Special care was devoted to the investigation of periodicities in the neighbourhood of 26 and 27 days .
Some rise in the periodogram for periods approximating to the time of synodic solar rotation may be expected if an appreciable number of sun-spots live through several rotations .
But apart from the fact that different circles of latitude have different periods of revolution , only a very broad band ( in the optical sense ) can be expected , as the number of rotations during which a sun-spot is observed very seldom exceeds three .
Such a succession could only produce a general rise between the periods of 25 and 29 days ; this rise does , indeed , seem to take place , but no definite periodicity has been discovered .
[ Note added December 12.\#151 ; Since writing the above , I have found strong evidence of the reality of a cycle having a periodic time of about 8*37 years .
This evidence , which seems to me convincing , is briefly as follows:\#151 ; If by 1905 .
] On Sun-spot Periodicities .
means of the table published by the Solar Physics Committee , giving the mean daily areas of spots for each year since 1832 , the average value for the eleven-year cycle in each year is calculated and deducted from each entry , we obtain a series of numbers which may be taken to represent the sun-spot variation with the eleven-year period eliminated .
This series shows decided maxima , which took place in the years 1836 , 1845 , 1853 , 1862 , and 1870 , the intervals being alternately nine and eight years .
The periodogram based on Wolf 's numbers for the total interval 1749 to 1900 shows a decided maximum for a periodicity of 8*25 years .
Provisionally accepting this period and taking the phase as obtained from the interval 1749 to 1826 , we may forecast the maxima for the subsequent interval .
We thus obtain : 1836*3 ; 1844*7 ; 1852*9 ; 1861*2 ; 1869*4 in almost exact agreement with the above .
Taking into account the apparent shift of phase which is noticed when the observations of the first and second portions of the complete interval are taken into account , the more exact time of the period seems to be 8*38 .
A further confirmation of this period may be obtained by means of an empirical criterion which I have found useful .
In a true periodicity , the main sine or cosine variation is in the spot cycles accompanied by the higher harmonic which has a period of half the length of the main period .
This harmonic is distinguished by a rise in the periodograph curve .
Such a rise where it exists gives increased probability to the real existence of the period considered .
In the case of fictitious periods , on the other hand , I have generally found that when the intensity of a certain period is exceptionally great , the whole of the accidental variations being as it were concentrated into it , the semi-period shows an exceptionally low intensity .
As regards the cycle under consideration , it is therefore significant that the periodogram has a maximum for the period of 4*125 years in both the intervals 1750 to 1826 and 1826 to 1900 .
In tracing backwards this new period as well as the one which is generally admitted to have a periodic time of a little over 11 years , the conclusion has been forced upon me that while the times of the maxima seem to occur with almost astronomical accuracy , the intensity of action is subject to great variations .
This conclusion I believe to give the key to the explanation of the great irregularities which are observed in the succession of sun-spot cycles .
While in the period 1750 to 1826 the eleven-year period does not appreciably affect the periodogram , it was not totally absent , for what there remains of it fits in well with the phase of the great eleven-year cycle which has been observed since 1826 .
I have approached this question without preconceived opinions , and my first impression on looking at the periodogram tended towards a denial of fixed periods extending over long ranges of time , Prof. A. Schuster .
[ Nov. 23 , but gradually the conviction has forced itself upon me that there are a number of perfectly definite periodicities , having the peculiarity that for a certain number of cycles they are effective and then cease to be active .
Their real existence is proved by the fact that whenever they reappear after a period of inactivity , they form , without change of phase , a continuation of the former periodicities .
The phenomenon reminds one to some extent of the beats of sound , but the evidence at present is against an explanation founded on the theory of interference of two nearly equal oscillations .
The periodogram for the total interval 1825 to 1900 shows a rise for a periodicity of about 13'5 years .
Looking at the records of sun-spot maxima since 1600 , we find three examples of two successive maxima being separated by an interval of about 13*5 years .
They are : 1626*0 to 1639*5 ; 1816*4 to 1829*9 ; and 1870*6 to 1883*9 .
Taking the interval between 1639*5 and 1816*4 , we find it to be 13 times 13*61 , while the interval between 1829 9 and 1870*6 is three times 13*57 .
This would tend to show that the maxima in question belong to a periodicity of about 13*57 years , which has extended right through two and a-half centuries .
It is specially to be noted that no observation previous to 1749 was used in determining the periodogram , so that the coincidence of the maxima as far back as 1626 is remarkable .
Without wishing to attach too great a weight to what may quite possibly be only a numerical accident , I think that as a confirmation of conclusions arrived at independently , the coincidence deserves notice .
The periodicities which so far have been traced with a considerable amount of probability were found to have periodic times of 4*81 , 8*38 , and 11*125 years .
There are indications that the first of these periods is somewhat less than the number given , and at one time I had put it down as low as 4*75 , but the above were actually the times directly determined , and I adopt them so as to avoid suspicion of having in any way helped by personal bias to improve the following remarkable relationship between the numbers .
That there was a connection between the numbers was only discovered after the first draft of my paper had been written out giving the above figures as highly probable periodicities .
Taking frequencies into consideration in place of periodic times , we are led to consider reciprocals , and thus find\#151 ; ( 11T25)-1 = 0*08989 ( 8*38)-1 = 0*11933 Sum = ( 4*78)-1 = 0*20922 Hence the sum of the frequencies of two of the periods agrees , within the possible errors , with the frequency of the third period .
But it is also found that the two first numbers are very nearly in the ratio 1905 .
] On Sun-spot ' Periodicities .
of three to four , so that we may also express the three periodic times as subperiods of 33375 years ; thus\#151 ; |x 33-375 = 11-125 |x 33-375 = 8-344 | x 33-375 = 4-768 How far this connection is accurate or approximate , it is impossible to say at present , but the fact that the three periods which have been traced with a considerable degree of certainty should also bear a remarkably simple relationship to each other is worthy of note .
If we accept a period twice as long as that given above , we might account for other periodicities of which at present the times are only approximately determined ; thus 4 x 66'75 would lead us to 13-34 , in fair agreement with the period of 13*57 years which has been mentioned above .
But for the present I do not wish to lay any stress on this .
I have confined the discussion to the statistical problem , but could not help giving some thought to the possibilities of explanation .
It is not difficult to form a theory which should account for the peculiarity of interruption in the succession of cycles which has been found to be characteristic of the sun-spot periods .
A reason may also be found for the first relationship between the cycles .
But the harmonic dependence on a long period is more difficult to account for .
]

