IV .
" On a Theorem relating to Polar Umbrae .
" By J. J. SYLVESTER , M.A. , F.R.S. Received April 27 , 1863 .
By polar umbree I mean such as obey in the strictest manner the polar law of sign , so that not only any two appositions or products of such umbrae derivable from one another by an interchange of two of their elements are to be considered each as the negative of the other , but also any such apposition or product becomes zero if the same element is found in it more than once .
Thus Sir W. Hamilton 's i , j , k are not polar umbrae , because although ijk= -jik kij , &amp;c. , ii , j , kk , instead of being nulls , are in the Calculus of Quaternions taken as unities* .
Let us now define any set arranged either in line or column of such umbral quantities to be multiplied by a corresponding set of actual quantities when each term of the one set is multiplied by the corresponding one of the other , and the sum taken of the products so obtained as in the ordinary case of the multiplication of the lines or columns of two determinants inter se .
Thus , ex. yr .
( a , b , cix , y , z ) , as also ( bsI/ is to mean the same product , viz. ax+ by +cz .
Again , imagine a rectangular ( square or oblong ) matrix of polar umbre , and that each line thereof is multiplied by the same line of actual quantities , the product of the products so obtained I call a Factorial of the Matrix .
I also call the product similarly obtained when the columns of the matrix are substituted for the lines , a Factorial of the same , but distinguish between the two by giving to one the name of a Transverse , the second of a Longitudinal Factorial of the matrix .
We are now in a position to enunciate the following remarkable theorem : The product of any longitudinal by any transverse factorial of the same polar umbral matrix is identically zero .
Ex. yr .
Let of be a matrix of polar umbrae , but x , y , z and also 4 , n actual quantities .
Then ( ax + b , cz ) ( dx + ey +fz ) is a transverse factorial , ( a +d/ ) ( b +en ) ( c +fT ) a longitudinal factorial of the above matrix , and by the theorem their product should be zero .
This is easily verified .
The two factorials expanded are respectively adx2 + bey2 + cfz + ( ae+ bd)xy + ( bf+ ce)yz+ ( af+ dc)zx , abcS3 + ( abf+ aec + dbc ) , ri + ( dec+ dbf+ aef)W2 + defr3 ; in their product the coefficient of x2T3 ==aboad=O , zxy3 =abcae + abebd= O , t242 " = abfad+ aecad+ dbcad= 0 , wyT2 = abfae + abfbd+ aecae + aecbd+ dbcae + dbcdb = aecbddbeae= aeebdaecbd= 0 , and so for all the other terms .
This is the fundamental theorem by aid of which I obtain the resultant of a lineo-linear system of equations in its most perfect form .
It is easy to obtain two different solutions , each of them unsymmetrical in respect of the data of the question ; the conversion and fusion of each of these into one and the same determinant , symmetrical in all its relations to the data , is effected instantaneously by a process derived from the above theorem .
In that particular application of it , the umbre involved each represent columns of actual quantities in number equal to the number of places in the width and length of the umbral matrix to which they belong , so that each coefficient in the product of a lateral by a longitudinal factorial represents an ordinary determinant made up of these columns , from which it is evident that the polar law of sign and nullity necessary for the truth of the theorem are satisfied in the case supposed .

