]\gt ; The odogram its Analogy .
By tTHUR s , F.F.S. ( Received November 23 , \mdash ; Read December 7 , 1905 .
) I ve recently applied the periodogram method to the investigation of several fluctuating quantities , and the experience thus gained has led me to llodify slightly original definition .
* Having always laid stress on the fact that the periodogram supplies by calculation the transformation which ectroscope instrumentally impresses on a luminous disturbance , I may now enter a little more closely into this optical , and thus lead up to whal I hope will ) the fmal definition .
Consider a parallel beam of falling on a grating , the reflected light colle ted at the focus of an telescope in the usual way .
For of calculation I assume that the grating considered is of a ticular type , which , in a former paper , I have called a simple Such a grating only gives two spectra of the first order .
If be the velocity at any ) of the incident beam , the displacement at the focus of the observing telescope is where nx .
( 1 ) In these equations denotes the focal of the teJescope , is the length of the lines ruled on the grating , the width of ruled space measured at right angles to the lines , gives the number of lines , and the between the direction of the optic axis of the observing telescope and the normal to the grating .
For the sake of shortness is written for .
The quantity denoted by is the wave-length of ] light which would have its first principal maximum at the focus of the telescope .
lt may be said to be the th towards which the telescope points , its strict definition is given by the relation where is the of incidence .
In order not to complicate needlessly the calculations , I shall assume that resolving power ufficient to ensure that at any ) oint of the spectrum 'Cambridge Phil. Soc. Trans vol. 18 , p. 107 .
'Phil .
Mag vol. 37 , p. 545 ( 1894 ) .
Periodogram and its Optical Analogy .
the vibrations are nearly eneous ; this involves that the average squares of the velocities are sensibly equal to the squares of the displacements multiplied by .
The average square of the velocity at the focus of the telescope is in that case\mdash ; where for we must put its average value .
This expression represents the measure of the intensity at the point considered .
Its line and surface integrals may be called the total linear intensity and the total intensity respectively .
In observin a spectrum , we associate with a ular wave-length all the which lies a straight line parallel to the rulings of the The distribution of along a vertical line for nearly eneous light takes according to the law , where being the vertical distance .
Multiplying by , and integrating from minus to plus infinity , the total intensity in a vertical line is found to be when the intensity at the central maximum is unity .
With the value for the central intensity previously found , we now obtain the total linear intensity associated with to be .
( 2 ) the variable , the expression for takes the form where is put for Write The mean of is then equal to the mean value of .
A such as that to which the above equations apply forms two spectra and absorbs part of the light ; we must now estimate what fraction of the incident beam is utilised to form the spectrum under consideration .
For this purpose we homogeneous light to fall on the grating , and put .
The mean value of 12 ) is then easily found to be .
By substitution into ( 2 ) we find that the total linear intensity in the central line is now .
( 3 ) Prof. A. Schuster .
[ Nov. 23 , To either side of the principal maximum the intensity varies according to the law , where is equal to representing a distance measured at right angles to the spectroscopic line .
The total energy measured in the focal plane of the telescope is obtained by multiplying ( 3 ) with , and integrating .
This gives If the incident light is rmal to the grating , its total energy is , the factor the fact that we ]lave taken the average square of the velocity which is half the square of the greatest velocity as the measure of intensity .
We conclude that is the fraction of light utilised to form the spectrum .
Taking account in ( 2 ) of this , we find that the type of spectroscope considered estimates the intensity of light passing through its central meridian as being , where for and their mean are to be substituted .
To obtain the total ) within a small angular distance , we must multiply by ; as , we find that the total energy within a range is If the total energy cf the incident on the grating is unity , the energy assigned by the ) rating to a is therefore finally\mdash ; .
( 4 ) In the application of the periodogral1l it is more convenient to take the time as the independent variable .
] ) } therefore\mdash ; , ( 5 ) ( 4 ) becomes equal to Leaving out the constant factor , I llow define to be the ordinate of the period .
The delinition differs from the previous one by the factol NT , which occurs in the denominato ' instead of its square .
The present definition not only justify by the close optical tnalogy which has now been formally , but also the resulting convenience .
I hnve previously shown that , in the absence of ] lomogeneous 1905 .
] The Periodogram its periodicities , the average of increases in proportion to the lime interval , NT , which occurs in the limits of the integrals for A and B. It follows that for such variations , the ordinate of the periodogram as at present defined is independent of the time limits chosen .
This is an advantage .
On the other hand , the former definition directly the amplitude of the periodic variation when it was of an absolutely homogeneous character .
For such eneous variation the present ordinate increases proportionally to the time interval chosen .
The optical explains the reason of this , and gives its justification .
When homogeneous light falls on an instrument of definite resolving power the in the central meridian does not by itself give sufficient indication of the intensity of the incident light .
It is only when correction has been made for the lateral that the true intensity can be deduced , the correction depending on the resolving power .
It is otherwise when the spectrum is continuous , for in that case the lost by lateral spreading is replaced by that which properly to the hbourino wave-lengths .
Hence , in this case , the intensity in the central meridian is a true measure of the intensity of the incident light .
It need hardly be pointed out how constant use is made of the fact that increased resolving power ( increased dispersion ) brings out the homogeneous lines of a spectrum by increasing their intensity that of the continuous background .
It is correspondingly one of the principal adyantages of the periodogram method that it gives a measure of the power necessary to isolate a true homogeneous period from the nlar fluctuations .
Light is thrown on parts of the previous investigaoion by a formula giveu by Lord for the intensity to be assigned to the homogeneous components of a disturbance .
If be the velocity at any point of a linear disturbance , so that the total intensity is Lord shows that the energy to be assigned to a range , where , is in which average intensity spread over a certain length may be estimated by taking and as lower and upper limits of the rals , and averaging the values obtained by a change of .
The energy per unit The Periodogrcorn its Optical Analogy .
would then be found on dividing the expression in ( 5 ) by L We arrive in this manner at equation ( 4 ) .
I might have confined myself to this simple deduction had I not wished to Iay stress on the equations for instrumental resolution .
This seemed all the more desirable because for absolutely homogeneous radiation a definition of the on the average intensity per unit length would fail .
If a simple periodicity exists , its alnplitnde may easily be derived from the ordinate of the periodogram , for , if be that ordinate , the amplitude is 2 In practical applications the function will yenerally lee given for successive intervals of the time .
The tJrals occurring in ( 5 ) are then placed by summations , unless a harmonic analyser is used .
It is most convenient to write in this case , ( 6 ) heal represents the values which takes at the successive times considered .
If we take to be equal to the total number of separate values of used in the calculations , we may put .
( 7 ) If a harmonic analyser ) used , and are two Fourier coefficients , where rep1esents the complete time }hich the Fourier analysis has applied .

