]\gt ; [ 19 ] .
On the of the in Linear Differential Operators .
E. B. ELLIOTT .
NI .
A. , Fellow of Queen 's College , Oxford .
by Professor SYLVESTER , F.R.S. Received June 5 , \mdash ; Read June 20 , 1889 .
CONTENTS .
Section .
1ntroduction . . . . . .
Definitions . . . . .
2 Symbolical basis of msthod of transformation . . . . . . .
3 , 4 , 5 Application to cases in which sum of degree and step is not less than unity 6 , 7 The special case of .
Transformation of . . . . .
8 , 9 , 10 Cases of . . . . . . . .
111 , 12 , Cases of . . . . .
14 , 15 II .
TERNARY OPERATORS .
Definitions and basis of method . . . . .
16 , 17 , 18 Proof and exemplification of formulae of cyclicaI 19 , 20 , 21 Transformation of . . . . . .
22 Operators free from first derivatives . . . . .
23 Transformation of . . . . .
24 1 .
The operators to be considered , include or involve all those which have presented themselves as annihilators and generators in recent theories of functional differential invariants , reciprocants , cyclicants , \amp ; c. The general form of the binary operators , operators whose arguments are the derivatives of one dependent with regard to one independent variable , which I propose first to consider , is adopted in accordance with that used in two remarkable papers by Major They are his operators in four elements .
The analogous ternary operators to which I subsequently devote attention , are distinct from his operators of six elements .
Their arguments are the partial derivatives of one of three variables , supposed connected by a single relation , with regard to the two others .
* ' The Theory of a Multilinear Partial Differential Operator with applications to the Theories of Invariatlts and Reciprocants ' London Math. Soc. Proc vol. 18 , 1887 , pp. 61-88 .
" " The AIgebra ot Multilinear Partial Differential Operators ' London Math. Soc. Proc vol. 19 , pp. 112-128 .
14.3.90 The only previous contribution , of which I am aware , to the subject of the reversion of MAcMAHON operators , is a paper by Professor L. J. ROGERS , which he obtains the operator reciprocal to ; 1 , 1 , and alludes to the self reciprocal property of V which is express with more precision in ( 38 ) below .
I. Binary 2 .
Let and be two variables connected by any relation .
Let denote , and , denote Let and be corresponding increments of and , so that by TAYLOR'S theorem ( 1 ) . .
, ( 2 ) the one expansion being a reversion of the other .
Let denote the coefficient of in the expansion of , i.e. , of in ascending integral powers of ; and the coefficient of in the expansion of , i. e. , of in ascending integral powers of It is supposed that is not fractional .
It need not , however , be positive .
Nor is it necessary to exclude the value zero , which , though somewhat special , will be seen to be of importance later .
Let be a positive or negative integer or zero , and and be any numerical quantities .
Denote by , . . . .
( 3 ) and by , . . . . .
( 4 ) the summations being , with regard to , which assumes in turn all integral values not less than the greater of the two and , so that , if , only symbok of differentiation with regard to all derivatives from onwards may occur , while , if , symbols of differentiation with regard to all derivatives may be present .
It is the operators and of which I propose to speak as MacMahon operators in and , respectively , dependent .
It will be seen upon reference to the first paper referred to above that they are the results of substitution in Major MAcMAlION'S operator for * ' ' Note on Conjugate Annihilators of Homogemeous and Isobaric Differential Equations ' Messenger of Mathematics , ' vol. 18 , pp. 153-158 .
VARIABLES IN CERTAIN LINEAR FERENTIAL RATORS .
of , . .
the one case , and of , . .
in the other .
MAcMAHON himself generally takes them as meanin , . .
, a fact which must not be forgotten in connecting his results with those to be here obtained .
The essential difference between the cases of and should be noticed at the utset .
Tn the former case , the complete set of coefficients appears in the operator .
In the latter , one or more of coefficients ( a number of them equal to the excess of over m ) is wanting at the beginning .
3 .
The aim in view is to any xcMahon operator in dependent as an operator or sum of operators of like form in dependent .
We need the linear expressions in , . .
which , when operating on any function of , . .
are equivalent to , .
perating on the equa ] function of , . .
The expressions in question I have obtained in the of a series of papers on Cyclicants , \amp ; c. The best form for present purposes is hardly there given to the conclusions .
It will therefore resuIt in a gain of clearness and no Ioss of brevity if in the present articlc the proof is given rather than the result quoted .
The same remark will apply to Article 17 below .
We may look upon , . . .
as a number of independent quantities , upon , . .
as determinate functions of these quantities , and upon and as two other quantities connected with one another and with , . .
by ( 1 ) or its equivalent ( 2 ) .
Give alone of all the quantites , . .
an variation .
Keep constant .
In virtue of ( 2 ) or its equivalent ( 1 ) will vary in consequence of the variation of .
Also , as some or all of , . .
are functions of , some or all of those quantities will vary .
Thus , from ( 2 ) we shall obtain @ and from ( 1 ) . .
* ' ' On the Linear Partial Differential Eqnations satisfied by Pure Ternary Reciprocants ' London Math. Soc. Proc vol. 18 , 1887 , pp. 142-164 .
Accordingl it follows that ; ( 5 ) and consequently , this being true for all values of , that if by aid of ( 1 ) this right hand member be expanded in ascending powers of , the coefficients of , .
are exactly the expressions for ] , .
Now , and is therefore the result of replacing each power of on the left , and therefore on the right , of ( 5 ) by the corresponding symbol .
It follows that the expression on the right of ( 5 ) may be taken as a symbolical representation of the equivalent operator to , i.e. , that , ( 6 ) where the meaning of the symbolisation on the right is that is to be replaced by ibs equivalent in terms of from ( 1 ) , that the differentiation with regard to is partial , that the product on the right is to be expanded as a sum of multiples of powers of and that then each power is to be replaced by the corresponding symbol of difl'erentiation 4 .
The proof of ( 6 ) is the same for all positive integral values ( including unity ) of Thus the means of transforming any differential operator whatever is obtained .
The rule according to which any linear operator is transformed may be very simply expressed .
Exactly companion to the symbolical notation for in an operator linear in , . .
is the notation for in an operator linear in , . .
Now , writing a linear operator A in the symbolical form , we learn by ( 6 ) that its equivalent in , . .
is obtained by multiplication by , expansion in of by ( 1 ) , and substitution for each power in the expanded result , of the corresponding VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
5 .
Now the symbolical form of any bIacMahon operator for which is very simple .
By ( 4 ) that of , is .
( 7 ) Thus in particular ( 8 ) and ( 9 ) ' the right hand members being supposed to be expanded in terms of by aid of ( 1 ) , and then to have each power of which occurs replaced by the corresponding W.e may of course write ( 7 ) , . . . .
( 10 ) so that in ( 8 ) and ( 9 ) we have involved all MacMahon operators in for which is not less than unity .
A reservation must for the moment be made of the case Exactly corresponding to ( 8 ) and ( 9 ) we have the symbolical forms , ( 11 ) ( 12 ) ' where the expansions on the right are to be in ascen powers of by ( 2 ) , and where in an expansion each is to be replaced by the corresponding 6 .
The transformation of for the cases at present under consideration of not less than unity is now immediate .
By Art .
4 the transformed form of the expansion in terms of of MR. Hi .
B. ELLIOTT ON THE INTERCHANGE OF THE considered as the symbolical form of an operator in dependent , is the expansion in terms of of considered as an operator in dependent .
In other words , by ( ] 1 ) and ( 9 ) , . . .
( 13 ) Again , the transformed form of the symboiical expansion powers of of is , by Art .
4 , the symbolical expansion in powers of of , of since in and the derivatives , and , .
are not arded as variables .
In other words , by ( 12 ) and ( 8 ) , . . . .
( 14 ) It is to be remarked that ( 13 ) and ( 14 ) are entirely in accord .
Either of them is produced from the other by the interchange of and and of and From ( 13 ) and ( 14 ) by aid of ( 10 ) we produce the more general equality of operators ( 15 ) which may be given the more symmetrical form , . . .
( 16 ) in which has to be positive .
In ( 16 ) are included two interesting classes of particular cases , viz. , . . .
( 17 ) and ( 1s ) Corresponding to each positive degree there are then two self-reciprocal operators .
The first is of positive character , being entirely unaltered in form by reciprocal ) , of conrso , generate from , bsolute ( .iprocants other absolute )rocants .
VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
interchange of and ; and the second of negative character , persisting in form but for a change of sign .
A complex self reciprocal operator can of course be found by taking the sum or difference of any two correlative operators ; e.g. , .
At greater length ( 17 ) and ( 18 ) are . .
, ( 17A ) and ( 18A ) In particular for we have .
( 19 ) and ( 20 ) Again gives us that . .
, . . . .
( 21 ) and , ( 22 ) are self reciprocal operators of positive and negative characters respectively .
7 .
As other examples of the important formulae of transformation ( 13 ) and ( 14 ) , let us write down cases corresponding to For we obtain , ( 23 ) . . . . .
, ( 24 ) nDCCCXC.\mdash ; A. which together are equivalent to ( 19 ) and ( 20 ) together .
From ( 23 ) it follows a homogeneous function of , . .
transforn ) into an isobaric function of , and that , and meaning degree and weight respectively , while from ( 24 ) follows the equivalent fact that an isobaric function transforms into a homogeneous one , and that From ( 19 ) follows the especially interesting fact that , if a function of , .
is isobaric in , .
upon considering the weight of to be , so also is the transformed function of , .
isobaric in the same sense and of the same weight in , .
Again the substitution in ( 13 ) and ( 14 ) produces for us ( 25 ) and , ( 26 ) where .
These two transformations have been obtained by Professor ROGERS ( see note to Art .
1 ) .
The second tells us that what he calls primary invariants in , .
have for their transforms what he calls secondary invariants in , .
We might now consider the results of putting and , 3 , . .
in ( 13 ) and ( 14 ) .
By this means he transformation of lineo-linear operators of two , three , \amp ; c. , steps is effected .
For the general case the results are .
( 27 ) Perhaps the most interesting fact to be deduced from ( 25 ) and ( 26 ) is the transfo1mation of , 1 ; 1 , 1 , the s.econd annihilator of projective reciprocants .
By subtraction of ) from ( 26 ) , or directly from ( 15 ) \mdash ; 1 , 1 ; 1 , VARIABLES IN CERTAJN LINEAR ) RENTIAL OPFRATORS .
i. e. , , or ( 29 ) Since we infer from this conclusion and its correlative that ( 80 ) is a self reciprocal operator of character .
The operator is one of considerable interest in connection with the theories of invariants and reciprocants .
( See , 'London Math. Soc. Proc vol. 18 , p. 75 ) .
It also follows that the sum of and the opel.ator on the ] of ( 30 ) is a self reciprocal operator of ) ositive character .
8 .
To complete our eory of the reversion of MacMahon operators , for which is not less than , we must consider the somewhat special and exactly correlative cases , and The operator is in accordance with the general ition of in ( 3 ) .
Thus the identity ( 6 ) may be written , by aid of ( 9 ) , , 0 ; 0 , , . . . . .
( 31 ) for any positive integral value , not excluding unity , of : which is .
in agreement with the general formula of transformation ( 13 ) .
On the other hand the general definition gives to no other meaning than zero .
So far then the operator is indeterminate in form .
An interpretation of it is now sought which shall make the case not exceptional to the general formula of transformation ( 14 ) .
To discover this pretation l us reverse the order of investigation and seek the operator in , which is equivalent to the operator in obtained by putting in the right hand member of ( 14 ) .
By ( 8 ) the symbolical form of is The equivalent operator in dependent has then for its symbolical form , as in ArticIe 4 , i. e. , Now , ( 32 ) where .
, so that the numerators , . .
, are a set of seminvariant protomorphs in , . .
Consequently the transformation in dependent is ; and it is accordingly this operator which has to be defined as ( 33 ) *Cf .
HAmMOND , ' London Math. Soc. Proc vol. 18 , p. 64 , note .
VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
that ( 14 ) may be regarded as holding the value as well as for hing values of It affords an instructive verification to conduct the investigation of the same transformation in the order of Article 6 .
9 .
For the case the two formulae of transformation , , 0 ; 0 , , 1 ; 0 , of the last article become respectively , ( 34 ) and .
( 35 ) By combination of these we have the equivalence , free from and . . .
( 36 ) In like manner , for any positive integral value of , 1 ; 0 , , 0 ; 0 , . . . . . . . . .
( 37 ) is an equivalence of operators which do not involve any lower symbols 01 operation than and respectively .
10 .
From ( 34 ) is easily derived in exact form the known fact that the annihilator V of } ) reciprocants is , when affected with a simple multiplier , self reciprocal .
We may write ( 34 ) , 1 ; 1 , So too , correlatively , , 1 ; 1 , .
But 1 , 1 ; 1 , by ( 18 ) or ( 20 ) ; . . . . .
so that , to use a familiar notation , is a self reciprocal operator of negative character .
11 It remains to consider operators in cases when .
In such cases the formulae of Arts .
5 and 6 have to be replaced by others .
The essential difference between them and the cases already considered lies in the fact that the lower limit of in ( 3 ) and ( 4 ) , and , therefore , in what replaces ( 7 ) , is now instead of , i.e. , is greater than , so that the coefficients in are no longer multiples of the complete set of coefficients in the expansion of , but of those coefficients with the exception of one or more at the beginning .
In the present article attention is confined to the case of , i.e. , Proceeding to write down the symbolical form of as in Art .
( 5 ) we see that the whole expansion from which we there started is present except the first term , Thus the symbolical form of is , ( 39 ) and that of is ( 40 ) the right-hand members standing for their expansions in powers of VARIABLES IN CRTAIN LINEAR DIFFERENTIAL OPERATORS .
In like manner and , standing for their expansions in terms of , are the symbolical forms of and As in Art .
( 6 ) the result of transforming { 1 , to its form in depsn dent is , then , symbolically , i. e. , ; whence , by aid of ( 40 ) , .
( 43 ) Once more the result of transforming is , in like manner , i.e. , ' i.e. , ; so that , by ( 39 ) and ( 40 ) , .
( 44 ) From ( 43 ) and ( 44 ) follows the more general identity , , 1 ; 1 , ; or , replacing by ( 45 ) In particular , , \mdash ; ( 46 ) 12 .
The value zero of is somewhat special in these cases of , just as in the more general cases already discussed .
So too , of course , is the conjugate value For ( 43 ) and ( 44 ) to hold for these special values of we must have , , 0 ; 0 , , 1 ; 0 , ( 47 ) , 0 ; 0 , Of these four equalities the first is a mere identity of two zero operators .
In fact to no other meaning than zero could be attached consistently with the general definition .
Thus the form of is left indeterminate .
The fourth of ( 47 ) becomes i. e. , , ( 48 ) of which the left-hand member is merely , and the right , the symbols of differentiation being tot , The remaining equalities , the second and third of ( 47 ) , now become the same but for an interchange of and .
Consequently there will be complete consistency if we define the at present undetermined operator as that which obeys the equation of transformation , 1 ; 0 , ( 49 ) VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
Now , proceeding exactly as in Art .
8 , it is seen that the symbolical form of an operator equal to this is , and that its expanded form is obtained by omitting the first term , and then putting in the general value ( 33 ) .
Thus , the operator which has to be defined as is . .
, ( 50 ) where is the generator defined in ( 32A ) .
13 .
Examples of important operators which occur among those transformed in the last two articles are before us in ( 48 ) and in the equality of ( 49 ) and ( 50 ) .
It is unnecessary to multiply particular instances as they can be deduced without number by giving particular integral values in ( 43 ) .
( 46 ) .
It is to be noticed that , excluding the special cases of and , one or other of the two equivalent operators will involve as coefficients those in a multinomial expansion of negative index .
Thus , for instance , a particular case of ( 43 ) , is at more length .
( 51 ) The transformation of the operator of ( 32A ) is an application of ( 49 ) .
Thus , 0 ; 1 , , 1 ; 1 , by ( 49 ) and ( 23 ) , .
( 52 ) MR. E. B. LIOTT ON THHi INTERCHANGE OF THE 14 .
Operators for which is negative still remain to be considered .
In particular , those of the type have still to be defined .
About the right definition of them there can , however , after articles 8 and 12 , be no doubt .
The general principle by means of which if is known , the form of is deduced is expressed by the rule\mdash ; " " Write symbolically , by putting for each , divide through by , reject all terms , if any now occur , which do not contain a positive power of as factor , and then for each write \ldquo ; Thus to accord with ( 33 ) and ( 50 ) the right operator to be defined as where is a negative integer , is ( 53 ) We now proceed to the transformation of where is a positive integer .
The symbolical form of is as in Arts .
6 and 11 .
( 54 ) The symbolical form of its transformation is , therefore , . . . .
( 55 ) This when expanded in terms of can involve no zero or tive powers .
For it is a sum of multiples of , only , since is a sum of multiples of , . .
only , and these when expressed in terms of are all free from Thus the coefficients of , . . .
1 , which would appear to occur in the above symbolical form of the transformation of are in reality absent ; and , consequently , , ( 56 ) the various terms on the right consisting of the parts with positive indices of from the corresponding terms of ( 55 ) .
VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
In like manner whose symbolical form is . .
, ( 57 ) a fortn proceeding by positive integral powers of , and therefore of , transforms into . . .
, ( 58 ) of which the terms in zero and negative powers of must , as before , disappear , leaving as the result transformation ( 59 ) By addition of times ( 56 ) to times ( 59 ) the transformation of the more general is at once deduced .
15 .
The forms taken by ( 56 ) and ( 59 ) for the case 1 , i.e. , , since and may be written , 1 ; 1 , ( 60 ) and ( 61 ) One or two particular cases of these formulae deserve mention .
The value zero of makes ( 60 ) an identity .
In ( 61 ) the substitution of the same value produces MR. ] .
B. ELLIOTT ON THE INTERCHANGE OF THE in verification of which we may notice that it is unaltered by interchange of and in virtue of ( 48 ) .
Another way of writing the result is to say that ( 62 ) is a self reciprocal operator of positive character .
We are now enabled to write ( 61 ) which becomes ( 60 ) upon interchanging and , replacing , and using ( 48 ) and the values for and , in terms of and .
Thus we have another verification of the consistency of our results .
II .
Ternary Operators .
16 .
Let be three variables connected by a relation of any form known or unknown .
Let denote respectively .
Let be any set of corresponding increments of .
They are connected by a single relation , which may be written in either of the forms . .
( 63 ) . .
, ( 64 ) . . .
( 65 ) Let be a positive integer , and let denote the coefficient of in ' when expanded in ascending products of positive integral powers of and , so that VARIABLES IN CERTAIN LINEAR NTIAL OPERATORS .
and in like manner write , ( 67 ) and .
( 68 ) We may include , if we please , the value zero of ; but the expansions of consist only of the single terms The operators to be considered and transformed are the following:\mdash ; , ( 69 ) ' ( 70 ) ; ( 71 ) $ where are any numerical quantities , a positive integer , positive integers or one or both zero , and quantities which take in succession all zero and positive integral values subject to Cases of negative , and of either or both less than \mdash ; 1 , which have been dealt with in the analogous theory of binary operators , will not be here considered .
The cases of zero , and of or equal to \mdash ; 1 , will not be entirely excluded , but will be only dealt with as far as their accordance with the results for positive and not negative needs no elaboration to make it clear .
Thus our field of investigation is rrower than in that of the analogous theory hitherto considered .
Were negative values of and admitted , the lower limit of in the operators ( 69 ) . .
( 71 ) would instead ofzero , and that of would be like manner .
Thus when we admit the value \mdash ; 1 of we must exclude the value of , and when the value of we must exclude the value of Let us now , ( 70 ) , ( 71 ) symbolica ] lie as follows:\mdash ; , ( 72 ) , ( 73 ) ( 74 ) i.e. , let us in any -operator symbolize by , in any -operator by , and in any -operator by We may in this way write ( 71 ) or ( 74 ) . .
, ( 75 ) where means the expansion in terms of and given in ( 65 ) , and where the symbolization denotes that the right-hand member is to be expanded in terms of and , and to have each product in its expansion replaced by the corresponding , in order to produce the operator in dependent which is represented by the notation on the left .
Thus in particular , assigning to different pairs in succession of the three parameters , , zero values , , . . . . .
( 76 ) , ( 77 ) , ( 78 ) while .
( 79 ) Precisely similar symbolical expressions to ( 75 ) . .
( 78 ) are , of course , assigned to the corresponding operators in and in dependent .
We have only cyclically to interchange once and twice respectively , and to regard the expressions on the right thus obtained as short ways of writing their expansions by aid of ( 63 ) and ( 64 ) in terms of and respectively .
17 .
In the present article the expression of each operative symbol , on a function of the derivatives of with regard to and , in terms of the operative symbols on the equivalent function of the derivatives of with regard to and , is investigated .
VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
If , as in the earlier part of the last article , are simultaneous increments of , we may look upon .
as a number of independent quantities ; upon , . . .
and , . . .
as determinate functions of these quantities ; upon as three quantities connected with one another , and with , by a relation of which ( 63 ) , ( 64 ) , and ) are equivalent forms .
Of the quantities , . .
let one alone receive an infinitesimal variation : also of , let and be kept constant so that receives a consequent variation .
Some or all of , . .
and some or all of , .
will also receive consequent variations .
From ( 63 ) we thus obtain ; from ( 64 ) @ ; and from ( 65 ) @ The three relations are identical .
Let us study the identity of the first and third We obtain from them that . . .
, ( 80 ) for all values of and ; and , consequently , that if by aid of ( 65 ) the right hand member be like the left , expanded in powers and products of powers of and , the coefficients of corresponding terms on the two sides will be equal .
In other words , each is the coefficient of the corresponding Now , in the equivalence of operators , each is the coefficient of It follows that in the expansion in terms of and of the right hand member of ( 80 ) the substitution for each of the corresponding exactly produces the expression for .
In other words , for each , the transform of , ( 81 ) where and its partial differential coefficient are to be replaced by their equivalents in terms of and by ( 65 ) , where the product is to be expanded in terms of and and where in the expanded result each product is to be replaced by the corresponding operative symbol By ( 77 ) we see that is thus replaced by a linear -operator of the form under consideration ; in fact that .
( 82 ) Since the and the of this operator are zero , the fact that is gives no difficulty as to the presence or absence of coefficients on the right like In precisely the same way , by giving variations to and in ( 63 ) and ( 64 ) instead of ( 63 ) and ( 65 ) , we might have obtained , ( 83 ) of which -operator the symbolical form is 18 .
The rules for transforming any linear -operator to its equivalent forms in and in dependent , are now very simply expressedjust as was the analogous rule in in Art .
4 .
Since the -operator or has for its equivalent -operator and for its equivalent -operator VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
these rules are merely\mdash ; To find the equivalent -operator to a given linear -operator , multiply its symbolical form by ; and to find its equivalent -operator , multiply that symbolical form .
The -operator thus obtained has of course to be expanded in terms of and by ( 64 ) , and the -operator in terms of and by ( 65 ) , before being intelligible except by means of ( 76 ) to ( 78 ) .
In verification let it be noticed that , since by first principles of the theory of partial differentiation the three sets of ratios : : : : , are equal , precisely the same results are obtained by cyclical interchange of and 19 .
Now as in ( 76 ) Its forms in and respectively are then , and Of these by two cyclic interchanges in ( 78 ) , and by ( 77 ) itself , respectively , the expressions are , and consequently .
( 84 ) In the same way the and transforms of are , and i. e. , , and CCCXC .
by the equalities of ratios at the end of the last article .
Accordingly .
( 85 ) And once more , precisely in the same way , and ' are equivalent operators in respectively dependent , so that also .
( 86 ) Of these sets of equalities ( 85 ) and ( 86 ) may in reality be deduced from ( 84 ) by cyclical interchanges of the variables and alteration of parameters .
The independent investigation above is justified by the verification it affords .
The general formulae of transformation , including ( 84 ) , ( 85 ) , ( 86 ) , follow from them by ( 79 ) , and are .
( 87 ) Included , it is interesting to notice that we have three distinct classes of self reproductive or cally persistent operators , of characters corresponding each to one of the cube roots of unity , viz. : ( 88 ) ( 89 ) ( 90 ) 20 .
Some of the simplest , and most important so far as actual experience goes , examples of the formulae now proved will be considered in what follows .
VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
The only lineo-linear operators , of the classes with which we are dealing , both of whose cyclical transformations are also lineo-linear , are found by putting and all equal to unity in the results of the last article .
Thus 1 , , 0 ; 1 , , 1 , 0 ; 1 , , ( 91 ) with the correlative equalities obtained by writing and respectively for , involve the aggregate of all such operators .
At length the equalities ( 91 ) are . . .
( 92 ) We thus learn that , if a function of the derivatives of with regard to and is homogeneous , the equivalent function of the derivatives of with regard to and is isobaric in second suffixes , while the equivalent function of the derivatives of with regard to and is isobaric in first suffixes ; and that , . . .
( 93 ) where the notation explains itself .
The oorrelative facts are , ( 94 ) and .
( 95 ) The same aggregate as is involved in ( 91 ) and its correlatives is also expressed b.y the facts that \mdash ; 1 , 1 , 1 ; 1 , , . . . .
( 96 ) , . . . . .
( 97 ) , . . .
( 98 ) obtained by giving the value 1 in ( 88 ) to ( 90 ) , are cyclical]y persistent lineo-linear operators of characters 1 , respectively .
If the operation be on a homogeneous and doubly isobaric function we are thus told that ( 99 ) are characteristics which persist after one cydical transformation but for the multipliers 1 , respectively , and after a second but for 1 , 21 .
The quadro-linear operators ( linear operators with coefficients quadratic in the derivatives ) both whose cyclic tran sformations are also quadro-linear , are obtained by giving to every one of the value 2 in the formulae of Art .
19 .
Their aggregate is involved in , 0 ; 2 , 1 , , 1 , 0 ; 2 , 1 , , ( 100 ) and its two correlatives in and same system is expressed by the.three cyclically persistent quadro-linear operators , 1 , 1 ; 2 , 1 , 1 , of character 1 , ( 101 ) , , , , . . . . .
( 102 ) .
( 103 ) Of these the first expanded to a few terms is , ( 104 ) where , . .
, and generally VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
The two imaginary cyclically persistent quadro-linear operators ( 102 ) and ( 103 ) are easil .
written out in like manner .
They commence with terms in , which it is to be observed are wanting from the above .
Once more by giving to each of the value 3 in Art .
19 , an aggregate is obtained of linear operators with coefficients of the third degree , whose transforms have both of them coefficients of the third degree also .
The aggregate may , as before , be considered involved in three cyclically persistent operators of the type , one of each character .
Similarly as to operators with coefficients of any higher degree .
22 .
Some of the most important linear operators which have been used in recent theories of functional invariants , cyclicants , \amp ; c. , have the property of persistence of degree in the derivatives after one cyclical transformation , but not after a second .
Such operators occur among those obtained by putting in ( 87 ) , ; ( 105 ) In particular , there are three classes of operators which have a property closely akin to that of persisting in form after a first cyclical transformation , being , in fact , only altered by the interchange of first and second suffixes : they are . . . .
( 106 ) . . .
( 107 ) . . .
( 108 ) It is to be noticed , in the case of the first of these , that the second cyclical transformation , which is of different degree from the first , is quite symmetrical in first and second suffixes .
Among the operators comprised in ( 106 ) occur the two , which I have called and two of the six form annihilators of projective cyclicants , viz. , * ' London Math. Soc. Proc vol. 20 , pp. 131-160 .
or in present notation and , 1 , 1 ; 1 , 1 , .
For the transformation of these we have , by putting 1 for each of and in ( 106 ) , , . . . .
( 109 ) of which right hand operator the expansion is . .
where the coefficients have meanings , as in Art .
21 .
Closely resembling , but distinct from and , are Mr. FORSYTH'S and i.e. , , 1 , 1 ; 1 , , 1 , 1 ; 1 , 1 , These are also transformed by means of the present article , but have not the property of companionship belonging to and .
In fact , by ( 105 ) , , 1 , 0 ; 1 , 1 , , ( 110 ) and , by ( 105 ) , with put for . . .
( 111 ) 23 .
The special importance of many operators in which the first derivatives do not occur is well known .
The form of such operators ( in dependent ) is symbolically As every such operator is a sum of multiples of complete operators so that their theory is implicitly involved in that above discussed , no attempt will be made here to develope it independently .
In the present article , however , an interesting class of cyclically persistent operators will be obtained , and a method of procedure in a much wider class of cases will be thus exemplified .
It is required to prove that the result of replacing each first derivative by zero in * See his Memoir " " A Clmss of Functional Invariants ' Phil. Trams A. , vol. 180 ( 1889 ) , pp. VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
is , but for a first de factor , an operator which persists in form after one and two cyclical interchanges of the variables .
Symbolically we have , if square brackets that in operator first .
are thus omitted , The transform of this operator is therefore , by Art .
18 , which , since and may be written and is consequently In exactly the same way the transform of the same operator is Thus we have the formula of transformation which may be written in a form even more clearly expressive of the cyclically persistent property , viz. , , l , 1 ; ( 112 ) A first case of this formula occurs in ( 96 ) .
A second is that the operator on the right of ( 109 ) is the sum of terms involving , and of an operator free from those derivatives which , when affected with the factor , has the persistent property .
The fact is interesting in its bearing on the theory of the transformation of and In fact ( 109 ) may be written From this and its two correlatives a ification of the formula ( 112 ) for the case of is readily obtained .
The first few terms of this cyclically persistent operator are .
( 113 ) With it compare .
the analogous operator ( 30 ) of the binary theory .
That , like this , is , of course , one of a whole class of persistent operators .
24 .
The transformation of the four form annihilators of pure cyclicants* will be now considered .
They are all operators free from first derivatives , and might as such be treated by the methods of the last article .
It seems preferable , however , to take them in connection with the formulae of transformation , ( 114 ) , ( 114A ) which are particular cases of ( 82 ) and ( 83 ) , or again of ( 87 ) with The forms of the four annihilators are ( 115 ) my papers in the ' London Math. Soc. Proc vol. 18 , pp. 142 , \amp ; c. ; vol. 19 , pp. 6 , \amp ; c. , and pp. 387 , \amp ; c. ; vol. 20 , pp. \amp ; c. VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
; ( 116 ) , 1 , 0 ; 1 , , 1 , 0 ; 1 , \mdash ; 1 , , 1 , 0 ; 2 , , 1 , 0 ; 1 , ( 117 ) ( 118 ) Thus may be written and the result of one cyclica ] intexchsnge in ( 114 ) may be written ; from which two sets of identities , by aid of the facts that : \mdash ; 1 : it follows at once that .
( 119 ) which , and its correlatives , ained by one and two cyclical transformations effect the transformation of and Again , we may write ( 114A ) and ( 1.14 ) in like manner From these it folows at once that \mdash ; By a cyclical interchange of the variables we have , also , ; and by a second cyclical interchange a third set of such equalities is obtained .
From the two sets that have been written out , upon subtraction we obtain , 1 , 0 ; 1 , \mdash ; , 1 ; 1 , VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
Now , from ( 91 ) and its correlative obtained by a cyclical ange , the second arts of these three equal operators are themselves equal .
Consequently , ( 120 ) which , and its correlatives , are the formulae for the transformation of and It is easy and very instructive to prove ( 120 ) directly from the symbolical expressions in ( 117 ) and ( 118 ) by the method of Art .
19 or 23 .
Of other important operators Mr. FORSYTII'S and ( ' Phil. Trans A. , vol. 180 , p. 74 ) should have their formulae of transformation noted .
They are the complete operators and of and are all but the first terms .
Thus their formulae of transformation are merely ( 114 ) and ( 114A ) themselves , i.e. , cyclically interchanging the variables once , , ( 121 ) .
( 122 ) [ 19 ] .
On the of the in Linear Differential Operators .
E. B. ELLIOTT .
NI .
A. , Fellow of Queen 's College , Oxford .
by Professor SYLVESTER , F.R.S. Received June 5 , \mdash ; Read June 20 , 1889 .
CONTENTS .
Section .
Introduction . . . . . . .
1 I. BINARY OPERATORS .
Definitions . . . . .
2 Symbolical basis of msthod of transformation . . . . . . .
3 , 4 , 5 Application to cases in which sum of degree and step is not less than unity 6 , 7 The special case of .
Transformation of . . . . .
8 , 9 , 10 Cases of . . . . . . . .
111 , 12 , Cases of . . . . .
14 , 15 II .
TERNARY OPERATORS .
Definitions and basis of method . . . . .
16 , 17 , 18 Proof and exemplification of formulae of cyclicaI 19 , 20 , 21 Transformation of . . . . . .
22 Operators free from first derivatives . . . . .
23 Transformation of . . . . .
24 1 .
The operators to be considered , include or involve all those which have presented themselves as annihilators and generators in recent theories of functional differential invariants , reciprocants , cyclicants , \amp ; c. The general form of the binary operators , operators whose arguments are the derivatives of one dependent with regard to one independent variable , which I propose first to consider , is adopted in accordance with that used in two remarkable papers by Major They are his operators in four elements .
The analogous ternary operators to which I subsequently devote attention , are distinct from his operators of six elements .
Their arguments are the partial derivatives of one of three variables , supposed connected by a single relation , with regard to the two others .
* ' The Theory of a Multilinear Partial Differential Operator with applications to the Theories of Invariatlts and Reciprocants ' London Math. Soc. Proc vol. 18 , 1887 , pp. 61-88 .
" " The AIgebra ot Multilinear Partial Differential Operators ' London Math. Soc. Proc vol. 19 , pp. 112-128 .
14.3.90 The only previous contribution , of which I am aware , to the subject of the reversion of MAcMAHON operators , is a paper by Professor L. J. ROGERS , which he obtains the operator reciprocal to ; 1 , 1 , and alludes to the self reciprocal property of V which is express with more precision in ( 38 ) below .
I. Binary 2 .
Let and be two variables connected by any relation .
Let denote , and , denote Let and be corresponding increments of and , so that by TAYLOR'S theorem ( 1 ) . .
, ( 2 ) the one expansion being a reversion of the other .
Let denote the coefficient of in the expansion of , i.e. , of in ascending integral powers of ; and the coefficient of in the expansion of , i. e. , of in ascending integral powers of It is supposed that is not fractional .
It need not , however , be positive .
Nor is it necessary to exclude the value zero , which , though somewhat special , will be seen to be of importance later .
Let be a positive or negative integer or zero , and and be any numerical quantities .
Denote by , . . . .
( 3 ) and by , . . . . .
( 4 ) the summations being , with regard to , which assumes in turn all integral values not less than the greater of the two and , so that , if , only symbok of differentiation with regard to all derivatives from onwards may occur , while , if , symbols of differentiation with regard to all derivatives may be present .
It is the operators and of which I propose to speak as MacMahon operators in and , respectively , dependent .
It will be seen upon reference to the first paper referred to above that they are the results of substitution in Major MAcMAlION'S operator for * ' ' Note on Conjugate Annihilators of Homogemeous and Isobaric Differential Equations ' Messenger of Mathematics , ' vol. 18 , pp. 153-158 .
VARIABLES IN CERTAIN LINEAR FERENTIAL ORATORS .
of , in the one case , and of , . .
in the other .
MAcMAHON himself generally takes them as meanin , . . . .
, a fact which must not be forgotten in connecting his results with those to be here obtained .
The essential difference between the cases of and should be noticed at the utset .
Tn the former case , the complete set of coefficients appears in the operator .
In the latter , one or more of coefficients ( a number of them equal to the excess of over m ) is wanting at the beginning .
3 .
The aim in view is to any xcMahon operator in dependent as an operator or sum of operators of like form in dependent .
We need the linea , expressions in , . .
which , when operating on any function of , . .
are equivalent to , .
perating on the equal function of , . .
The expressions in question I have obtained in the of a series of papers on Cyclicants , \amp ; c. The best form for present purposes is hardly there given to the conclusions .
It will therefore result in a gain of clearness and no loss of brevity if in the present article the proof is given rather than the result quoted .
The same remark will apply to Article 17 below .
We may look upon , . . .
as a number of independent quantities , upon , .
as determinate functions of these quantities , and upon and as two other quantities connected with one another and with , . .
by ( 1 ) or its equivalent ( 2 ) .
Give alone of all the quantites , .
an ilffinitesimal variation .
Keep constant .
In virtue of ( 2 ) or its equivalent ( 1 ) will vary in consequence of the variation of .
Also , as some or all of , . .
are functions of , some or all of those quantities will vary .
Thus , from ( 2 ) we shall obtain @ and from ( 1 ) . .
' ' On the Linear Partial Differential Eqnations satisfied by Pure Ternary Reciprocants ' London Math. Soc. Proc vol. , 1887 , pp. 142-164 .
Accordingl it follows that ; ( 5 ) and consequently , this being true for all values of , that if by aid of ( 1 ) this right hand member be expanded in ascending powers of , the coefficients of , .
are exactly the expressions for ] , .
Now , and is therefore the result of replacing each power of on the left , and therefore on the right , of ( 5 ) by the corresponding symbol .
It follows that the expression on the right of ( 5 ) may be taken as a symbolical representation of the equivalent operator to , i.e. , that ( 6 ) where the meaning of the symbolisation on the right is that is to be replaced by ibs equivalent in terms of from ( 1 ) , that the differentiation with regard to is partial , that the product on the right is to be expanded as a sum of multiples of powers of and that then each power is to be replaced by the corresponding symbol of difl'erentiation 4 .
The proof of ( 6 ) is the same for all positive integral values ( including unity ) of Thus the means of transforming any differential operator whatever is obtained .
The rule according to which any linear operator is transformed may be very simply expressed .
Exactly companion to the symbolical notation for in an operator linear in , . .
is the notation for in an operator linear in , . .
Now , writing a linear operator A in the symbolical form , we learn by ( 6 ) that its equivalent in , . .
is obtained by multiplication by , expansion in of by ( 1 ) , and substitution for each power in the expanded result , of the corresponding VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
5 .
Now the symbolical form of any bIacMahon operator for which is very simple .
By ( 4 ) that of , is .
( 7 ) Thus in particular ( 8 ) and ( 9 ) ' the right hand members being supposed to be expanded in terms of by aid of ( 1 ) , and then to have each power of which occurs replaced by the corresponding W.e may of course write ( 7 ) , . . . .
( 10 ) so that in ( 8 ) and ( 9 ) we have involved all MacMahon operators in for which is not less than unity .
A reservation must for the moment be made of the case Exactly corresponding to ( 8 ) and ( 9 ) we have the symbolical forms , ( 11 ) ( 12 ) ' where the expansions on the right are to be in ascen powers of by ( 2 ) , and where in an expansion each is to be replaced by the corresponding 6 .
The transformation of for the cases at present under consideration of not less than unity is now immediate .
By Art .
4 the transformed form of the expansion in terms of of MR. Hi .
B. ELLIOTT ON THE INTERCHANGE OF THE considered as the symbolical form of an operator in dependent , is the expansion in terms of of considered as an operator in dependent .
In other words , by ( ] 1 ) and ( 9 ) , . . .
( 13 ) Again , the transformed form of the symboiical expansion powers of of is , by Art .
4 , the symbolical expansion in powers of of , of since in and the derivatives , and , .
are not arded as variables .
In other words , by ( 12 ) and ( 8 ) , . . . .
( 14 ) It is to be remarked that ( 13 ) and ( 14 ) are entirely in accord .
Either of them is produced from the other by the interchange of and and of and From ( 13 ) and ( 14 ) by aid of ( 10 ) we produce the more general equality of operators ( 15 ) which may be given the more symmetrical form , . . .
( 16 ) in which has to be positive .
In ( 16 ) are included two interesting classes of particular cases , viz. , . . .
( 17 ) and ( 18 ) Corresponding to each positive degree there are then two self-reciprocal operators .
The first is of positive character , being entirely unaltered in form by reciprocal ) , of conrso , generate from , bsolute ( .iprocants other absolute )rocants .
VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
interchange of and ; and the second of negative character , persisting in form but for a change of sign .
A complex self reciprocal operator can of course be found by taking the sum or difference of any two correlative operators ; e.g. , .
At greater length ( 17 ) and ( 18 ) are . .
, ( 17A ) and ( 18A ) In particular for we have .
( 19 ) and ( 20 ) Again gives us that . .
, . . . .
( 21 ) and , ( 22 ) are self reciprocal operators of positive and negative characters respectively .
7 .
As other examples of the important formulae of transformation ( 13 ) and ( 14 ) , let us write down cases corresponding to For we obtain , ( 23 ) . . . . .
, ( 24 ) nDCCCXC.\mdash ; A. which together are equivalent to ( 19 ) and ( 20 ) together .
From ( 23 ) it follows a homogeneous function of , . .
transforn ) into an isobaric function of , and that , and meaning degree and weight respectively , while from ( 24 ) follows the equivalent fact that an isobaric function transforms into a homogeneous one , and that From ( 19 ) follows the especially interesting fact that , if a function of , .
is isobaric in , .
upon considering the weight of to be , so also is the transformed function of , .
isobaric in the same sense and of the same weight in , .
Again the substitution in ( 13 ) and ( 14 ) produces for us ( 25 ) and , ( 26 ) where .
These two transformations have been obtained by Professor ROGERS ( see note to Art .
1 ) .
The second tells us that what he calls primary invariants in , .
have for their transforms what he calls secondary invariants in , .
We might now consider the results of putting and , 3 , . .
in ( 13 ) and ( 14 ) .
By this means he transformation of lineo-linear operators of two , three , \amp ; c. , steps is effected .
For the general case the results are .
( 27 ) Perhaps the most interesting fact to be deduced from ( 25 ) and ( 26 ) is the transfo1mation of , 1 ; 1 , 1 , the s.econd annihilator of projective reciprocants .
By subtraction of ) from ( 26 ) , or directly from ( 15 ) \mdash ; 1 , 1 ; 1 , VARIABLES IN CERTAJN LINEAR ) RENTIAL OPFRATORS .
i. e. , , or ( 29 ) Since we infer from this conclusion and its correlative that ( 80 ) is a self reciprocal operator of character .
The operator is one of considerable interest in connection with the theories of invariants and reciprocants .
( See , 'London Math. Soc. Proc vol. 18 , p. 75 ) .
It also follows that the sum of and the opel.ator on the ] of ( 30 ) is a self reciprocal operator of ) ositive character .
8 .
To complete our eory of the reversion of MacMahon operators , for which is not less than , we must consider the somewhat special and exactly correlative cases , and The operator is in accordance with the general ition of in ( 3 ) .
Thus the identity ( 6 ) may be written , by aid of ( 9 ) , , 0 ; 0 , , . . . . .
( 31 ) for any positive integral value , not excluding unity , of : which is .
in agreement with the general formula of transformation ( 13 ) .
On the other hand the general definition gives to no other meaning than zero .
So far then the operator is indeterminate in form .
An interpretation of it is now sought which shall make the case not exceptional to the general formula of transformation ( 14 ) .
To discover this pretation l us reverse the order of investigation and seek the operator in , which is equivalent to the operator in obtained by putting in the right hand member of ( 14 ) .
By ( 8 ) the symbolical form of is The equivalent operator in dependent has then for its symbolical form , as in ArticIe 4 , i. e. , Now , ( 32 ) where .
, so that the numerators , . .
, are a set of seminvariant protomorphs in , . .
Consequently the transformation in dependent is ; and it is accordingly this operator which has to be defined as ( 33 ) *Cf .
HAmMOND , ' London Math. Soc. Proc vol. 18 , p. 64 , note .
VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
that ( 14 ) may be regarded as holding the value as well as for hing values of It affords an instructive verification to conduct the investigation of the same transformation in the order of Article 6 .
9 .
For the case the two formulae of transformation , , 0 ; 0 , , 1 ; 0 , of the last article become respectively , ( 34 ) and .
( 35 ) By combination of these we have the equivalence , free from and . . .
( 36 ) In like manner , for any positive integral value of , 1 ; 0 , , 0 ; 0 , . . . . . . . . .
( 37 ) is an equivalence of operators which do not involve any lower symbols 01 operation than and respectively .
10 .
From ( 34 ) is easily derived in exact form the known fact that the annihilator V of } ) reciprocants is , when affected with a simple multiplier , self reciprocal .
We may write ( 34 ) , 1 ; 1 , So too , correlatively , , 1 ; 1 , .
But 1 , 1 ; 1 , by ( 18 ) or ( 20 ) ; . . . . .
so that , to use a familiar notation , is a self reciprocal operator of negative character .
11 It remains to consider operators in cases when .
In such cases the formulae of Arts .
5 and 6 have to be replaced by others .
The essential difference between them and the cases already considered lies in the fact that the lower limit of in ( 3 ) and ( 4 ) , and , therefore , in what replaces ( 7 ) , is now instead of , i.e. , is greater than , so that the coefficients in are no longer multiples of the complete set of coefficients in the expansion of , but of those coefficients with the exception of one or more at the beginning .
In the present article attention is confined to the case of , i.e. , Proceeding to write down the symbolical form of as in Art .
( 5 ) we see that the whole expansion from which we there started is present except the first term , Thus the symbolical form of is , ( 39 ) and that of is ( 40 ) the right-hand members standing for their expansions in powers of VARIABLES IN CRTAIN LINEAR DIFFERENTIAL OPERATORS .
In like manner and , standing for their expansions in terms of , are the symbolical forms of and As in Art .
( 6 ) the result of transforming { 1 , to its form in depsn dent is , then , symbolically , i. e. , ; whence , by aid of ( 40 ) , .
( 43 ) Once more the result of transforming is , in like manner , i.e. , ' i.e. , ; so that , by ( 39 ) and ( 40 ) , .
( 44 ) From ( 43 ) and ( 44 ) follows the more general identity , , 1 ; 1 , ; or , replacing by ( 45 ) In particular , , \mdash ; ( 46 ) 12 .
The value zero of is somewhat special in these cases of , just as in the more general cases already discussed .
So too , of course , is the conjugate value For ( 43 ) and ( 44 ) to hold for these special values of we must have , , 0 ; 0 , , 1 ; 0 , ( 47 ) 1 , 0 ; 1 , , 0 ; 0 , Of these four equalities the first is a mere identity of two zero operators .
In fact to no other meaning than zero could be attached consistently with the general definition .
Thus the form of is left indeterminate .
The fourth of ( 47 ) becomes i. e. , , ( 48 ) of which the left-hand member is merely , and the right , the symbols of differentiation being tot , The remaining equalities , the second and third of ( 47 ) , now become the same but for an interchange of and .
Consequently there will be complete consistency if we define the at present undetermined operator as that which obeys the equation of transformation , 1 ; 0 , ( 49 ) VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
Now , proceeding exactly as in Art .
8 , it is seen that the symbolical form of an operator equal to this is , and that its expanded form is obtained by omitting the first term , and then putting in the general value ( 33 ) .
Thus , the operator which has to be defined as is . .
, ( 50 ) where is the generator defined in ( 32A ) .
13 .
Examples of important operators which occur among those transformed in the last two articles are before us in ( 48 ) and in the equality of ( 49 ) and ( 50 ) .
It is unnecessary to multiply particular instances as they can be deduced without number by giving particular integral values in ( 43 ) .
( 46 ) .
It is to be noticed that , excluding the special cases of and , one or other of the two equivalent operators will involve as coefficients those in a multinomial expansion of negative index .
Thus , for instance , a particular case of ( 43 ) , is at more length .
( 51 ) The transformation of the operator of ( 32A ) is an application of ( 49 ) .
Thus , 0 ; 1 , , 1 ; 1 , by ( 49 ) and ( 23 ) , .
( 52 ) MR. E. B. LIOTT ON THHi INTERCHANGE OF THE 14 .
Operators for which is negative still remain to be considered .
In particular , those of the type have still to be defined .
About the right definition of them there can , however , after articles 8 and 12 , be no doubt .
The general principle by means of which if is known , the form of is deduced is expressed by the rule\mdash ; " " Write symbolically , by putting for each , divide through by , reject all terms , if any now occur , which do not contain a positive power of as factor , and then for each write \ldquo ; Thus to accord with ( 33 ) and ( 50 ) the right operator to be defined as where is a negative integer , is ( 53 ) We now proceed to the transformation of where is a positive integer .
The symbolical form of is as in Arts .
6 and 11 .
( 54 ) The symbolical form of its transformation is , therefore , . . . .
( 55 ) This when expanded in terms of can involve no zero or tive powers .
For it is a sum of multiples of , only , since is a sum of multiples of , . .
only , and these when expressed in terms of are all free from Thus the coefficients of , . . .
1 , which would appear to occur in the above symbolical form of the transformation of are in reality absent ; and , consequently , , ( 56 ) the various terms on the right consisting of the parts with positive indices of from the corresponding terms of ( 55 ) .
VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
In like manner whose symbolical form is . .
, ( 57 ) a fortn proceeding by positive integral powers of , and therefore of , transforms into . . .
, ( 58 ) of which the terms in zero and negative powers of must , as before , disappear , leaving as the result transformation ( 59 ) By addition of times ( 56 ) to times ( 59 ) the transformation of the more general is at once deduced .
15 .
The forms taken by ( 56 ) and ( 59 ) for the case 1 , i.e. , , since and may be written , 1 ; 1 , ( 60 ) and ( 61 ) One or two particular cases of these formulae deserve mention .
The value zero of makes ( 60 ) an identity .
In ( 61 ) the substitution of the same value produces MR. ] .
B. ELLIOTT ON THE INTERCHANGE OF THE in verification of which we may notice that it is unaltered by interchange of and in virtue of ( 48 ) .
Another way of writing the result is to say that ( 62 ) is a self reciprocal operator of positive character .
We are now enabled to write ( 61 ) which becomes ( 60 ) upon interchanging and , replacing , and using ( 48 ) and the values for and , in terms of and .
Thus we have another verification of the consistency of our results .
II .
Ternary Operators .
16 .
Let be three variables connected by a relation of any form known or unknown .
Let denote respectively .
Let be any set of corresponding increments of .
They are connected by a single relation , which may be written in either of the forms . .
( 63 ) . .
, ( 64 ) . . .
( 65 ) Let be a positive integer , and let denote the coefficient of in ' when expanded in ascending products of positive integral powers of and , so that VARIABLES IN CERTAIN LINEAR NTIAL OPERATORS .
and in like manner write , ( 67 ) and .
( 68 ) We may include , if we please , the value zero of ; but the expansions of consist only of the single terms The operators to be considered and transformed are the following:\mdash ; , ( 69 ) ' ( 70 ) ; ( 71 ) $ where are any numerical quantities , a positive integer , positive integers or one or both zero , and quantities which take in succession all zero and positive integral values subject to Cases of negative , and of either or both less than \mdash ; 1 , which have been dealt with in the analogous theory of binary operators , will not be here considered .
The cases of zero , and of or equal to \mdash ; 1 , will not be entirely excluded , but will be only dealt with as far as their accordance with the results for positive and not negative needs no elaboration to make it clear .
Thus our field of investigation is rrower than in that of the analogous theory hitherto considered .
Were negative values of and admitted , the lower limit of in the operators ( 69 ) . .
( 71 ) would instead ofzero , and that of would be like manner .
Thus when we admit the value \mdash ; 1 of we must exclude the value of , and when the value of we must exclude the value of Let us now , ( 70 ) , ( 71 ) symbolica ] lie as follows:\mdash ; , ( 72 ) , ( 73 ) ( 74 ) i.e. , let us in any -operator symbolize by , in any -operator by , and in any -operator by We may in this way write ( 71 ) or ( 74 ) . .
, ( 75 ) where means the expansion in terms of and given in ( 65 ) , and where the symbolization denotes that the right-hand member is to be expanded in terms of and , and to have each product in its expansion replaced by the corresponding , in order to produce the operator in dependent which is represented by the notation on the left .
Thus in particular , assigning to different pairs in succession of the three parameters , , zero values , , . . . . .
( 76 ) , ( 77 ) , ( 78 ) while .
( 79 ) Precisely similar symbolical expressions to ( 75 ) . .
( 78 ) are , of course , assigned to the corresponding operators in and in dependent .
We have only cyclically to interchange once and twice respectively , and to regard the expressions on the right thus obtained as short ways of writing their expansions by aid of ( 63 ) and ( 64 ) in terms of and respectively .
17 .
In the present article the expression of each operative symbol , on a function of the derivatives of with regard to and , in terms of the operative symbols on the equivalent function of the derivatives of with regard to and , is investigated .
VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
If , as in the earlier part of the last article , are simultaneous increments of , we may look upon .
as a number of independent quantities ; upon , . . .
and , . . .
as determinate functions of these quantities ; upon as three quantities connected with one another , and with , by a relation of which ( 63 ) , ( 64 ) , and ) are equivalent forms .
Of the quantities , . .
let one alone receive an infinitesimal variation : also of , let and be kept constant so that receives a consequent variation .
Some or all of , . .
and some or all of , .
will also receive consequent variations .
From ( 63 ) we thus obtain ; from ( 64 ) @ ; and from ( 65 ) @ The three relations are identical .
Let us study the identity of the first and third We obtain from them that . . .
, ( 80 ) for all values of and ; and , consequently , that if by aid of ( 65 ) the right hand member be like the left , expanded in powers and products of powers of and , the coefficients of corresponding terms on the two sides will be equal .
In other words , each is the coefficient of the corresponding Now , in the equivalence of operators , each is the coefficient of It follows that in the expansion in terms of and of the right hand member of ( 80 ) the substitution for each of the corresponding exactly produces the expression for .
In other words , for each , the transform of , ( 81 ) where and its partial differential coefficient are to be replaced by their equivalents in terms of and by ( 65 ) , where the product is to be expanded in terms of and and where in the expanded result each product is to be replaced by the corresponding operative symbol By ( 77 ) we see that is thus replaced by a linear -operator of the form under consideration ; in fact that .
( 82 ) Since the and the of this operator are zero , the fact that is gives no difficulty as to the presence or absence of coefficients on the right like In precisely the same way , by giving variations to and in ( 63 ) and ( 64 ) instead of ( 63 ) and ( 65 ) , we might have obtained , ( 83 ) of which -operator the symbolical form is 18 .
The rules for transforming any linear -operator to its equivalent forms in and in dependent , are now very simply expressedjust as was the analogous rule in in Art .
4 .
Since the -operator or has for its equivalent -operator and for its equivalent -operator VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
these rules are merely\mdash ; To find the equivalent -operator to a given linear -operator , multiply its symbolical form by ; and to find its equivalent -operator , multiply that symbolical form .
The -operator thus obtained has of course to be expanded in terms of and by ( 64 ) , and the -operator in terms of and by ( 65 ) , before being intelligible except by means of ( 76 ) to ( 78 ) .
In verification let it be noticed that , since by first principles of the theory of partial differentiation the three sets of ratios : : : : , are equal , precisely the same results are obtained by cyclical interchange of and 19 .
Now as in ( 76 ) Its forms in and respectively are then , and Of these by two cyclic interchanges in ( 78 ) , and by ( 77 ) itself , respectively , the expressions are , and consequently .
( 84 ) In the same way the and transforms of are , and i. e. , , and CCCXC .
by the equalities of ratios at the end of the last article .
Accordingly .
( 85 ) And once more , precisely in the same way , and ' are equivalent operators in respectively dependent , so that also .
( 86 ) Of these sets of equalities ( 85 ) and ( 86 ) may in reality be deduced from ( 84 ) by cyclical interchanges of the variables and alteration of parameters .
The independent investigation above is justified by the verification it affords .
The general formulae of transformation , including ( 84 ) , ( 85 ) , ( 86 ) , follow from them by ( 79 ) , and are .
( 87 ) Included , it is interesting to notice that we have three distinct classes of self reproductive or cally persistent operators , of characters corresponding each to one of the cube roots of unity , viz. : ( 88 ) ( 89 ) ( 90 ) 20 .
Some of the simplest , and most important so far as actual experience goes , examples of the formulae now proved will be considered in what follows .
VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
The only lineo-linear operators , of the classes with which we are dealing , both of whose cyclical transformations are also lineo-linear , are found by putting and all equal to unity in the results of the last article .
Thus 1 , , 0 ; 1 , , 1 , 0 ; 1 , , ( 91 ) with the correlative equalities obtained by writing and respectively for , involve the aggregate of all such operators .
At length the equalities ( 91 ) are . . .
( 92 ) We thus learn that , if a function of the derivatives of with regard to and is homogeneous , the equivalent function of the derivatives of with regard to and is isobaric in second suffixes , while the equivalent function of the derivatives of with regard to and is isobaric in first suffixes ; and that , . . .
( 93 ) where the notation explains itself .
The oorrelative facts are , ( 94 ) and .
( 95 ) The same aggregate as is involved in ( 91 ) and its correlatives is also expressed b.y the facts that \mdash ; 1 , 1 , 1 ; 1 , , . . . .
( 96 ) , . . . . .
( 97 ) , . . .
( 98 ) obtained by giving the value 1 in ( 88 ) to ( 90 ) , are cyclical]y persistent lineo-linear operators of characters 1 , respectively .
If the operation be on a homogeneous and doubly isobaric function we are thus told that ( 99 ) are characteristics which persist after one cydical transformation but for the multipliers 1 , respectively , and after a second but for 1 , 21 .
The quadro-linear operators ( linear operators with coefficients quadratic in the derivatives ) both whose cyclic tran sformations are also quadro-linear , are obtained by giving to every one of the value 2 in the formulae of Art .
19 .
Their aggregate is involved in , 0 ; 2 , 1 , , 1 , 0 ; 2 , 1 , , ( 100 ) and its two correlatives in and same system is expressed by the.three cyclically persistent quadro-linear operators , 1 , 1 ; 2 , 1 , 1 , of character 1 , ( 101 ) , , , , . . . . .
( 102 ) .
( 103 ) Of these the first expanded to a few terms is , ( 104 ) where , . .
, and generally VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
The two imaginary cyclically persistent quadro-linear operators ( 102 ) and ( 103 ) are easil .
written out in like manner .
They commence with terms in , which it is to be observed are wanting from the above .
Once more by giving to each of the value 3 in Art .
19 , an aggregate is obtained of linear operators with coefficients of the third degree , whose transforms have both of them coefficients of the third degree also .
The aggregate may , as before , be considered involved in three cyclically persistent operators of the type , one of each character .
Similarly as to operators with coefficients of any higher degree .
22 .
Some of the most important linear operators which have been used in recent theories of functional invariants , cyclicants , \amp ; c. , have the property of persistence of degree in the derivatives after one cyclical transformation , but not after a second .
Such operators occur among those obtained by putting in ( 87 ) , ; ( 105 ) In particular , there are three classes of operators which have a property closely akin to that of persisting in form after a first cyclical transformation , being , in fact , only altered by the interchange of first and second suffixes : they are . . . .
( 106 ) . . .
( 107 ) . . .
( 108 ) It is to be noticed , in the case of the first of these , that the second cyclical transformation , which is of different degree from the first , is quite symmetrical in first and second suffixes .
Among the operators comprised in ( 106 ) occur the two , which I have called and two of the six form annihilators of projective cyclicants , viz. , * ' London Math. Soc. Proc vol. 20 , pp. 131-160 .
or in present notation and , 1 , 1 ; 1 , 1 , .
For the transformation of these we have , by putting 1 for each of and in ( 106 ) , , . . . .
( 109 ) of which right hand operator the expansion is . .
where the coefficients have meanings , as in Art .
21 .
Closely resembling , but distinct from and , are Mr. FORSYTH'S and i.e. , , 1 , 1 ; 1 , , 1 , 1 ; 1 , 1 , These are also transformed by means of the present article , but have not the property of companionship belonging to and .
In fact , by ( 105 ) , , 1 , 0 ; 1 , 1 , , ( 110 ) and , by ( 105 ) , with put for . . .
( 111 ) 23 .
The special importance of many operators in which the first derivatives do not occur is well known .
The form of such operators ( in dependent ) is symbolically As every such operator is a sum of multiples of complete operators so that their theory is implicitly involved in that above discussed , no attempt will be made here to develope it independently .
In the present article , however , an interesting class of cyclically persistent operators will be obtained , and a method of procedure in a much wider class of cases will be thus exemplified .
It is required to prove that the result of replacing each first derivative by zero in * See his Memoir " " A Clmss of Functional Invariants ' Phil. Trams A. , vol. 180 ( 1889 ) , pp. VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
is , but for a first de factor , an operator which persists in form after one and two cyclical interchanges of the variables .
Symbolically we have , if square brackets that in operator first .
are thus omitted , The transform of this operator is therefore , by Art .
18 , which , since and may be written and is consequently In exactly the same way the transform of the same operator is Thus we have the formula of transformation which may be written in a form even more clearly expressive of the cyclically persistent property , viz. , , l , 1 ; ( 112 ) A first case of this formula occurs in ( 96 ) .
A second is that the operator on the right of ( 109 ) is the sum of terms involving , and of an operator free from those derivatives which , when affected with the factor , has the persistent property .
The fact is interesting in its bearing on the theory of the transformation of and In fact ( 109 ) may be written From this and its two correlatives a ification of the formula ( 112 ) for the case of is readily obtained .
The first few terms of this cyclically persistent operator are .
( 113 ) With it compare .
the analogous operator ( 30 ) of the binary theory .
That , like this , is , of course , one of a whole class of persistent operators .
24 .
The transformation of the four form annihilators of pure cyclicants* will be now considered .
They are all operators free from first derivatives , and might as such be treated by the methods of the last article .
It seems preferable , however , to take them in connection with the formulae of transformation , ( 114 ) , ( 114A ) which are particular cases of ( 82 ) and ( 83 ) , or again of ( 87 ) with The forms of the four annihilators are ( 115 ) my papers in the ' London Math. Soc. Proc vol. 18 , pp. 142 , \amp ; c. ; vol. 19 , pp. 6 , \amp ; c. , and pp. 387 , \amp ; c. ; vol. 20 , pp. \amp ; c. VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
; ( 116 ) , 1 , 0 ; 1 , , 1 , 0 ; 1 , \mdash ; 1 , , 1 , 0 ; 2 , , 1 , 0 ; 1 , ( 117 ) ( 118 ) Thus may be written and the result of one cyclica ] intexchsnge in ( 114 ) may be written ; from which two sets of identities , by aid of the facts that : \mdash ; 1 : it follows at once that .
( 119 ) which , and its correlatives , ained by one and two cyclical transformations effect the transformation of and Again , we may write ( 114A ) and ( 1.14 ) in like manner From these it folows at once that \mdash ; By a cyclical interchange of the variables we have , also , ; and by a second cyclical interchange a third set of such equalities is obtained .
From the two sets that have been written out , upon subtraction we obtain , 1 , 0 ; 1 , \mdash ; , 1 ; 1 , VARIABLES IN CERTAIN LINEAR DIFFERENTIAL OPERATORS .
Now , from ( 91 ) and its correlative obtained by a cyclical ange , the second arts of these three equal operators are themselves equal .
Consequently , ( 120 ) which , and its correlatives , are the formulae for the transformation of and It is easy and very instructive to prove ( 120 ) directly from the symbolical expressions in ( 117 ) and ( 118 ) by the method of Art .
19 or 23 .
Of other important operators Mr. FORSYTII'S and ( ' Phil. Trans A. , vol. 180 , p. 74 ) should have their formulae of transformation noted .
They are the complete operators and of and are all but the first terms .
Thus their formulae of transformation are merely ( 114 ) and ( 114A ) themselves , i.e. , cyclically interchanging the variables once , , ( 121 ) .
( 122 )

