I. " On the Orders and Genera of Quadratic Forms containing more than three Indeterminates .
" By H. T. STEPHEN SMITHj M.A. , F.R.S. , Savilian Professor of Geometry in the University of Oxford .
Received March 22 , 1864 .
'Let us represent by f , a homogeneous form or quantic of any order containing n indeterminates ; by ( a( ' ) ) , a square matrix of order n ; by ( a ( ) , its ith derived matrix , i. e. the matrix of order r In =-=I , the conji |n-i stituents of which are the minor determinants of order i of the matrix ( M(1 ) ) ; and lastly , byf , a form of any order containing I indeterminates , the coefficients of which depend on the coefficients of fi .
When fA is transformed by ( a(1 ) ) , let fi be transformed by ( a(t ) ) ; if , after division or multiplication by a power of the modulus of transformationi , the metamorphic of fi depends on the metamorphic of f. , in the same way in whichfi depends onf1 , f is said to he a concomiitant of the ith species of f. Thus : a concomitant of the 1st species is a covariant ; a concomitant of the ( n-l)th species is a contravariant ; if n=2 there are only covariants ; if n=3 there are only covariants and contravariants ; but if n &gt; 3 , there will exist in general concomitants of the intermediate species .
There is anl obvious difference between covarianlts and contravariants on the one hand , and the intermediate concomitants on the other .
The number of indeterminates in a covariant or contravariant is the same as in its primitive ; in an intermediate concomitant , the num ber of indeterminates is always greater than in its primitive .
Again , to every metamorphic of a covarianit or contravariant , there corresponds a metamorphic of its primitive ; whereas , in the case of a concomitanit of the intermediate order i , a metamorphic of the primitive will correspond , not to every metamorplhic of the con-comitant , but only to such metamorphics as result from transformations the matrices of which are the ith derived matrices of matrices of order n. It is also obvious that , besides the n1 species of concomitanice here defined , there are , whenl n is &gt; 3 , an infinite number of other species of concomitance of the same general nature .
For from any derived matrix we may form another derived matrix , and so on continually ; and to every such process of derivation a distiniet species of colncomitance will correspond .
The notion of intermediate concomitance appears likely to be of use in many researches ; in what follows , it is employed to obtain a definition of the ordinal and generic characters of quadratic forms containing more than 3 indeterminates .
( The case of quadratic forms containing 3 indeterminates has been considered by Eisenstein in h-is memoir , " Never Theoreme des h6heren Arithmetik , " Crelle , vol. xxxv .
pp. 121 and 125 .
) Let P~=n q=92(1 fi vI A( ' ) Xp Cq p=l q=1P , q represent a quadratic form of n indeterminates ; let ( A( ' ) ) be the symmetrical matrix of this form , and ( A( ) ) the ith derived matrix of ( A( ' ) ) ( A(i ) ) will also be a symmetrical matrix , and the quadratic form p=I q=T i f=Y. .
LA ( A ) ' q=l2l P , ,q Pq will be a concomitant of the ith species of A. It is immaterial what principle of arrangement is adopted in writing the quadratic matrix ( A( ) ) , and the transforming matrix ( a($ ) ) ; provided only that the arrangement be the same in the two matrices , and that in each matrix it be the same in height and in depth .
For example , if f=atl X2 + a2 X+ a3 X+ a4X ' + 26l 1 2+ 2 263 t1 x4 +r 264 x+ 265 x,2 4+ 266 x,3 x be a quadratic form containing four indetermilnates , the form f= ( _~a1 a2 ) x2 ?
( 62_a1 a3 ) X2~ ?
( 63-a1 a4 ) X3 ?
2 ( b4-2 a3 ) x2 + ( 6 _ a , a4 ) X2 + ( 62 a , Ct ) X6 +2(61Z2-al 4)X1X + 2(6163-a163 ) X2 X2 2(6 4-a2 3 ) X1 X4-2(61 ba-a2 63 ) X1 X6 2(63 62 -al 64 ) XI X2 + 2(63 63 -al 66 ) X2 X3 ? ?
2 ( 61 64 -a3 62 ) Xl X4 2(61 65662 64 ) Xl XI5 -2(b2 ba 3 b4 ) Xl XG + 2(62 b3 Cl b6 ) X-2 X3 + 2(62 64-a3 63 ) X2 X4 2 ( 61 6 -63 64 ) X_X 42 ( 2 b6 3 b3 ) X2 XI,2 ( b1 b6 -b2 ba ) X3 XL + 2(63 65-a4 61 ) X3 X5 + 2(63 b6-a4 62 ) X3 X ? ?
2(64 65-a2 66 ) X4 X2(4 b6-a3 65 ) N4 X6 ?
2(b6 6.-a4 64 ) X5 X6 is the concomitant of the secolnd species off .
The n-1I forms defined by the formula ( A ) , of which the first is the form A itself , and the last the contravariant off , , we shall term thefundamentat concomitants of fi ; in conitradistinction to those other quadratic concomitants ( infinite in number ) of which the matrices are the symmetrical matrices that may be derived , by a multiplicate derivation , from ( A(1 ) ) ... .
Passing to the arithmetical theory of quadratic forms-i .
e. supposing that the constituents of ( A( ' ) ) are integral numbers , we shall designate by V1 , V2 , ... V the greatest common divisors ( taken positively ) of the minors of different orders of the matrix ( A(1 ) ) , so that , in particular , V1 is the greatest common divisor of its constituents , and Un is the absolute value of its determinant , here supposed to be different from zero .
By the primary divisor of a quadratic form we shall understand the greatest common divisor of the coefficients of the squares and double rectangles in the quadratic form ; by the secondary divisor we shall understand the greatest common divisor of the coefficients of the squares and of the rectangles ; so that the primary divisor is equal to , or is half of , the secondary divisor , according as the quadratic form ( to use the phraseology of Gauss ) is derived from a form properly or improperly primitive .
It will be seeln that V1 , V2 ... .
Vn-l are the primary divisors of the forms fi ' f2 .
fnA 1 respe-tively .
We now consider the totality of arithmetical quadratic forms , containing n indeterminates , and having a given index of inertia , and a given determinant .
If a quadratic form be reduced to a sum of squares by any linear transformationi , the number of positive and of negative squares is the same , whatever be the real transformation by which the reduction is effected .
For the index of inertia we may take the number of the positive squares ; it is equal to the number of continuations of sign in a series of ascending principal minors of the matrix of the quadratic form , the series commencing with unity , i. e. with a minor of order 0 , and each minor being so taken as to contain that which precedes it in the series ( see Professor Sylvester " On Formule connected with Sturm 's Theorem , " Phil. Trans. vol. cxliii .
p. 481 ) .
The distribution of these forms into Orders depends on the following principle : " Two forms belong to the same order when the primary and secondary divisors of their corresponding concomitants are identical .
" Since , as has been just pointed out , there are , beside the fundamental concomitants , an infinite number of other concomitants , it is important to know whether , in order to obtain the distribution inlto orders , it is , or is not , necessary to consider those other concomitants .
With regard to the primary divisors , it can be shown that it is unnecessary to consider any concomitants other than the fundamental ones ; i. e. it can be shown that the equality of the primary divisors of the corresponding fundamental concomitants of two quadratic forms , implies the equality of the primary divisors of all their corresponding concomitants .
And it is probable ( but it seems difficult to prove ) that the same thing is true for the secondary divisors also .
Confining our attention ( in the next place ) to the forms contained in any given order , we proceed to indicate the principle from which the subdivision of that order into genera is deducible .
If F. be any quadratic form containing r indeterminates , and F , be its concomitant of the second species , we have the identical equation F(1 -2 , ** r)X F](yl , Y21 ... Yr)-6 yk I2( F2{Z1 S2 * .
* SC.-k=1 dV * B ) F21 XI ) *2 X r } y1'~~ , ,,. .
in which the symbol F , ( 2 : y2 ' ) indicates that the determinants ( x prime dividing V ?
i ? ?
2L ; we infer from the identity ( B ) that the V ' Ve-1 numbers prime to 3i , which call be represented by Oi , are either all quadratic residues of ~i , or all non-quadratic residues of Bj .
In the former case we attribute to fi the particular character + I ; in the latter the particular character -1 .
If V1= 1 , i. e. if the form fA itself do not admit of any primary divisor beside unity ( which is the only important case ) , the product ( v Vn-YLz1x(7V'I.Vn-2 x. 'Vn-I 'Vn-2 lVn-2 'Vn-3 is equal to.n ; whence , inasmuch as every prime that divides un also divides -Y v , it appears that a primitive quadratic form will always have Vn-1 one particular character , at least with respect to every uneven prime dividing its determinant , and will have more than one if the uneven prime divide more than one of the quotients ?
i * Vi Vi-i The subdivision of an order into genera can now be effected by assigning to the same genus all those forms whose particular characters coincide .
But it remains to consider whether the above enumeration of particular characters is complete .
It is evident that we might apply the theorem ( B ) to other concomitants besides those included in the fundamental system ; and it might appear as if in this manner we could obtain other particular characters besides those which we have given .
But it can be shown that such other particular characters are implicitly contained in ours ; i. e. it can be shown that two quadratic forms , which coincide in respect of the particular characters deducible from their fundamental concomitants , will also coincide in respect of the particular characters deducible from any other concomitant .
Again , it will be found that if the determinant be uneven , there are no particular characters with respect to 4 or 8 .
For this case , therefore , our enumeratiGn is complete .
But when the determinant is even , besides the , particular characters arising from its uneven prime divisors , there may also be particular characters with regard to 4 or 8 .
There is no difficulty in enumerating these particular characters ; nevertheless we suppress the enumeration here , not only because it would require a detailed distinction of cases , but also because there appears to be some difficulty in showing that the characters with regard to 4 or 8 , which may arise from the excluded concomitants , are virtually included in those which arise from the concomitants of the fundamental set .
principle of arrangement is adopted in writing the quadratic matrix ( A( ) ) , and the transforming matrix ( a( ) ) ; provided only that the arrangement be the same in the two matrices , and that in each matrix it be the same in height and in depth .
For example , if f= al x , + a2 x2 + a3 x+ a42 + 2b x1 2+ 262 x1 3+ 263 xi x4 + 2b4 x2 x3 +42 66 wx xx be a quadratic form containing four indeterminates , the form f , = ( 6b2-a a2 ) X2 + ( 6b-a1 a3 ) X2 + ( b-a1 a ) X3 + ( 6b-2 a , ) X2 + ( 62--a2 a ) X+ ( 62-a a ) X2 + 2(6 , 62-a b4 ) X , X2 + 2(6 , 63 , -a b , ) X2 X , -2(b b4-a b2 ) X X4 2(6 , 6 , 63 ) X1 X , -2(6b2 b -3 64 ) X1 X , + 2(62 3a b66 ) X,2 X3 + 2(62 6b-3 b6 ) X2 X4 2(6 b6 6-3 b4 ) X2 , X -2(6 , b6-63 3 ) X X2 , 2(b , 66-62 b6 ) X3 X , + 2(63 6Z-a4 bl ) X3 X5 + 2(63 b , -a,4 6 ) X3 X , + 2(6,4 6-a2 b6 ) X4 X5 2(6 , 6 , -3 b6 ) X , X6 + 2(6b 66-a4 b4 ) X5 X6 is the concomitant of the second species off .
The n--1 forms defined by the formula ( A ) , of which the first is the formf , itself , and the last the contravariant off , we shall term thefundamental concomitants of fi ; in contradistinction to those other quadratic concomitants ( infinite in number ) of which the matrices are the symmetrical matrices that may be derived , by a multiplicate derivation , from ( A( ) ) ... .
Passing to the arithmetical theory of quadratic forms-i .
e. supposing that the constituents of ( A( ' ) ) are integral numbers , we shall designate by VI , V2 , .
Vs the greatest common divisors ( taken positively ) of the minors of different orders of the matrix ( A( ' ) ) , so that , in particular , V , is the greatest common divisor of its constituents , and V , is the absolute value of its determinant , here supposed to be different from zero .
By the primary divisor of a quadratic form we shall understand the greatest common divisor of the coefficients of the squares and double rectangles in the quadratic form ; by the secondary divisor we shall understand the greatest common divisor of the coefficients of the squares and of the rectangles ; so that the primary divisor is equal to , or is half of , the secondary divisor , according as the quadratic form ( to use the phraseology of Gauss ) is derived from a form properly or improperly primitive .
It will be seen that V , , V ... .
V. -i are the primary divisors of the forms f , f2 .
* whatever be the real transformation by which the reduction is effected , For the index of inertia we may take the number of the positive squares ; it is equal to the number of continuations of sign in a series of ascending principal minors of the matrix of the quadratic form , the series commencing with unity , i. e. with a minor of order 0 , and each minor being so taken as to contain that which precedes it in the series ( see Professor Sylvester " On Formulae connected with Sturm 's Theorem , " Phil. Trans. vol. cxliii .
p. 481 ) .
The distribution of these forms into Orders depends on the following principle:"Two forms belong to the same order when the primary and secondary divisors of their corresponding concomitants are identical .
" Since , as has been just pointed out , there are , beside the fundamental concomitants , an infinite number of other concomitants , it is important to know whether , in order to obtain the distribution into orders , it is , or is not , necessary to consider those other concomitants .
With regard to the primary divisors , it can be shown that it is unnecessary to consider any concomitants other than the fundamental ones ; i. e. it can be shown that the equality of the primary divisors of the corresponding fundamental concomitants of two quadratic forms , implies the equality of the primary divisors of all their corresponding concomitants .
And it is probable ( but it seems difficult to prove ) that the same thing is true for the secondary divisors also .
Confining our attention ( in the next place ) to the forms contained in any given order , we proceed to indicate the principle from which the subdivision of that order into genera is deducible .
If F1 be any quadratic form containing r indeterminates , and F , be its concomitant of the second species , we have the identical equation kdx Fl(xl , , .
)X F , ( ( Y1 , Y2 .
** *)-i[ ?
Y s ^dF ] l'B F/ ( , ' ik=1 r..* B)\Y1 Y2 ) * * prime dividing Vil ?
?-- ; we infer from the identity ( B ) that the Vi Vs-i numbers prime to Zi , which can be represented by Oi , are either all quadratic residues of ai , or all non-quadratic residues of 6i .
In the former case we attribute to ] the particular character ( - ) + 1 ; in the latter the particular character =1 .
If = 1 , i. e. if the form f , itself do not admit of any primary divisor beside unity ( which is the only important case ) , the product( Vnx~ V-l XVv X ... .
Vn-l Vn-2/ \vn-2 Vn-3/ is equal to V ; whence , inasmuch as every prime that divides Va also Vn-l divides -_VZ , it appears that a primitive quadratic form will always have Vn-i one particular character , at least with respect to every uneven prime dividing its determinant , and will have more than one if the uneven prime divide more than one of the quotients Vi+1 * .
Vi Vi-t The subdivision of an order into genera can now be effected by assigning to the same genus all those forms whose particular characters coincide .
But it remains to consider whether the above enumeration of particular characters is complete .
It is evident that we might apply the theorem ( B ) to other concomitants besides those included in the fundamental system ; and it might appear as if in this manner we could obtain other particular characters besides those which we have given .
But it can be shown that such other particular characters are implicitly contained in ours ; i. e. it can be shown that two quadratic forms , which coincide in respect of the particular characters deducible from their fundamental concomitants , will also coincide in respect of the particular characters deducible from any other concomitant .
Again , it will be found that if the determinant be uneven , there are no particular characters with respect to 4 or 8 .
For this case , therefore , our enumeration is complete , But when the determinant is even , besides the particular characters arising from its uneven prime divisors , there may also be particular characters with regard to 4 or 8 .
There is no difficulty in enumerating these particular characters ; nevertheless we suppress the enumeration here , not only because it would require a detailed distinction of cases , but also because there appears to be some difficulty in showing that the characters with regard to 4 or 8 , which may arise from the excluded concomitants , are virtually included in those which arise from the concomitants of the fundamental set .

