II .
A Method for determining the Number of impossible Roots in adfected AEquations .
By Mr. George Campbell .
LEMMA I. IN every adfected quadratick AEquation ax2Bx+ A= 0 , whose Roots are real , a fourth Part of the Square of the Coefficient of the second Term is greater than the Rectangle under the Coefficient of the first Term and the absolute Number or 1/ 4 B2~a~ A ; and vice versa if 1/ 4 B2~a~ A , the Roots of the AEquation ax2Bx+A= 0 , will be real .
But if 1/ 4 B2~a~ A , the Roots will be impossible .
This is evident from the Roots of the AEquation being 1/ 2 B+~ 1/ 4 B2a~ A/ a , 1/ 2 B~ 1/ 4 B2a~ A/ a. LEMMA II .
Whatever be the Number of impossible Roots in the AEquation xnBxn1+Cxn2Dxn3+ &amp;c. ~dx3~cx2~bx~A= 0 , there are just as many in the AEquation Axnbxn1+cxn2dxn3+ &amp;c. ~Dx3~Cx2~Bx~1= 0 .
For the Roots of the last AEquation are the Reciprocals of those of the first , as is evident from common Algebra .
Let the Roots of the biquadratick AEquation x4Bx3+ Cx2Dx+A=0 be a , b , c , d , whereof let c , d be impossible , then the Roots of the AEquation Ax4Dx3+Cx2Bx+1=0 will be 1/ a , 1/ b , 1/ c , 1/ d , and therefore two of them to wit 1/ c , 1/ d impossible .
LEMMA III .
In every AEquation xnBxn1+Cxn2Dxn3+Exn4 &amp;c. ~ex4~dx3~cx2~ bx~A= 0 , all whose Roots are real , if each Term be multiplied by the Index of x in that Term , and each Product be divided by x , the resulting AEquation nxn1n1Bxn2+n2Cxn3n3Dxn4+n4Exn5 &amp;c. ~4ex3 ~3dx2~2cx~b=0 shall have all its Roots real .
Thus if all the Roots of the AEquation x4Bx3+Cx2Dx+A=0 be real , then all the Roots of the AEquation 4x33Bx2+2CxD=0 will also be real .
This Lemma doth not hold conversly , for there are an Infinity of Cases where all the Roots of the AEquation nxn1n1Bxn2+ n2Cxn3n3Dxn4+ &amp;c. ~3dx2~ 2cx~b=0 are real , at the same Time some or perhaps all the Roots of the AEquation xnBxn1+ Cxn2Dxn3+ &amp;c. ~dx3~cx2~bx~A=0 are impossible : But whatever be the Number of impossible Roots in the AEquation nxn1n1Bxn2+ n2Cxn3 &amp;c. ~2cx~b= 0 , there are at least as many in the AEquation xnBxn1+ Cxn2 &amp;c. ~cx2~bx~A= 0 .
Thus all the Roots of the AEquation 4x33Bx2+2CxD=0 may be real , and yet two or perhaps all the four Roots of the AEquation x4Bx3+Cx2Dx+A=0 may be impossible , but if two of the Roots of the AEquation 4x33Bx2+2CxD=0 be impossible , there must be at least two impossible Roots in the AEquation x4Bx3+Cx2Dx+A= 0 .
All this hath been demonstrated by Algebraical Writers , particularly by Mr. Reyneau in his Analyse Demontre , and is easily made evident by the Method of the Maxima and Minima .
COROLAR Y. Let all the Roots of the AEquation xnBxn1+Cxn2Dxn3+Exn4Fxn5+ &amp;c. ~fx5~ex4~dx3~cx2~bx~ A=0 be real , and by this Lemma all the Roots of the AEquation nxn1n1Bxn2+n2Cxn3n3Dxn4+n4Exn5n5Fxn6+ &amp;c. ~5fx4~4ex3~3dx2~2cx~b=0 will be real , and therefore ( by the same Lemma ) all the Roots of the AEquation n~n1xn2n1~ n2Bxn3+n2~n3Cxn4n3~ n4Dxn5+n4~n5Exn6n5~n6 Fxn7+ &amp;c. ~ 20 fx3~12ex2~6dx~2c=0 or ( dividing all by 2 ) of n~n 1/ 2 xn2n1~ n 2/ 2 Bxn3+n2~n 3/ 2 Cxn4 &amp;c. ~ 10 fx3~6ex2~3dx~c=0 will be real .
After the same Manner all the Roots of the AEquation n~n 1/ 2 ~n 2/ 3 xn3n1~n 2/ 2 ~n 3/ 3 Bxn4+n2~n 3/ 2 ~n 4/ 3 Cxn5 &amp;c. ~ 10 fx2~4ex~d=0 will be real ; and thus we may descend until we arrive at the quadratick AEquation n~n 1/ 2 x2n1Bx+C= 0 .
The same AEquations do ascend thus n~n 1/ 2 x2n1Bx+ C= 0 , n~n 1/ 2 ~n 2/ 3 x3n1~n 2/ 2 Bx2+ n2CxD= 0 , n~n 1/ 2 ~n 2/ 3 ~n 3/ 4 x4 n1~n 2/ 2 ~n 3/ 3 Bx3+n2~n 3/ 2 ~ Cx2n3Dx+E= 0 , n~n 1/ 2 ~n 2/ 3 ~ n 3/ 4 ~n 4/ 5 x5n1~n 2/ 2 ~n 3/ 3 ~n 4/ 4 Bx4+n2~n 3/ 2 ~n 4/ 3 Cx3n3~n 4/ 2 Dx2+n4ExF= 0 , and so on .
Let M represent any of the Coefficients of the AEquation xnBxn1+Cxn2Dxn3+Exn4 &amp;c. ~A= 0 , and let LN be the adjacent Coefficients , let M be the Exponent of the Coefficient M : By the Exponent of a Coefficient I mean the Number which expresses the Place which it hath among the Coefficients , thus if M represent the Coefficient E ( and therefore L=D and N= F ) then m= 4 .
It will be easy to see , that , amongst the foregoing ascending AEquations , that which hath its absolute Number N will be n~n 1/ 2 ~n 2/ 3 ~ &amp;c. n m/ m +1xm+1n1~n 2/ 2 ~ &amp;c. n m/ m Bxm+ n2~ &amp;c. n m/ m 1Cxm1 &amp;c. ~nm+1~ n m/ 2 Lx2~nmMx~N= 0 , all whose Roots are real when all the Roots of the AEquation xnBxn1+Cxn~2 &amp;c. ~A=0 are real .
Let N=F and therefore M= E , L=D and m= 4 , then that of the ascending AEquations whose absolute Number is F , will be n~n 1/ 2 ~n 2/ 3 ~ n 3/ 4 ~n 4/ 5 x5n1~n 2/ 2 ~n 3/ 3 ~n 4/ 4 Bx4+n2~n 3/ 2 ~n 4/ 3 Cx3n3~n 4/ 2 Dx2+n4ExF= 0 .
2 PROPOSITION I. Let xnBxn1+Cxn2Dxn3+ Exn4 &amp;c. ~ex4~dx3~cx2~bx~A=0 be an AEquation of any Dimensions all whose Roots are real , let M be any Coefficient of this AEquation , L , N the adjacent Coefficients , and m the Exponent of M. Then the Square of any Coefficient M multiplied by the Fraction m~n m/ m +1~nm+1 will always exceed the Rectangle under the adjacent Coefficients L~ N. Thus in the AEquation x4Bx3+ Cx2Dx+A= 0 , where n= 4 , making M=C and therefore L= B , N= D , and m= 2 , then 2~4 2/ 2 +1~42+1~C2 or 4/ 9 C2 will exceed B~D providing all the Roots of the AEquation be real .
Because ( by Lem .
3 .
) the Roots of the quadratick AEquation n~n 1/ 2 x2n1Bx+C= 0 , are real , therefore ( by Lem .
1 .
) 1/ 4 n1~2~B2 must be greater than n~n 1/ 2 ~C and ( dividing both by n~n 1/ 2 ~n 1/ 2 n~B2 greater than 1~ C. Therefore in the AEquation xnBxn1+Cxn2Dxn3+ &amp;c. ~A=0 of the n Degree , all whose Roots are real , the Square of B the Coefficient of the second Term , multiplied by the Fraction n 1/ 2 n is greater than 1~C the Rectangle under the adjacent Coefficients .
But ( by Lem .
2 .
) all the Roots of the AEquation Ax~bxn1+cx~2 &amp;c. ~Cx2~Bx~ 1=0 or ( dividing by A ) of x~ b/ A x~1+ c/ A xn2 &amp;c. ~ C/ A x2~ B/ A x~ 1/ A =0 are real , therefore ( from what hath been just now said ) n 1/ 2 n~b 2/ A ~ must be greater than 1~ c/ A and consequently n 1/ 2 n~b2 greater than c~ A. Therefore in an AEquation x~Bxn1+Cx~~ &amp;c. ~cx2~bx~A= 0 , of the n Degree , all whose Roots are real , the Square of the Coefficient of x multiplied by the Fraction n 1/ 2 n is greater than the Rectangle under the Coefficient of x2 and the absolute Number .
But by Cor. Lem .
3 .
all the Roots of the AEquation n~n 1/ 2 ~n 2/ 3 ~ &amp;c. ~n m/ m +1xm+1n1~n 2/ 2 ~ &amp;c. ~n m/ m Bxm+n2~ &amp;c. n m/ m 1Cxm1 &amp;c. ~nm+1~n m/ 2 ~Lx2~ nmMx~N=0 are real , therefore ( seeing this AEquation is of the m+1 Degree ) the Square of nm~M multiplied by the Fraction m+1 1/ 2 ~m+1 will be greater than the Rectangle under nm+1~ n m/ 2 ~L and N , that is m/ 2 ~m+1~nm~2~ M2 will be greater than nm+1~n m/ 2 ~L~N and therefore ( dividing both by nm+1~n m/ 2 ~ m~n m/ m +1~nm+1~M2 greater than L~ N. COROLAR Y. Make a Series of Fractions n/ 1 , n 1/ 2 , n 2/ 3 , n 3/ 4 , &amp;c. unto 1/ n whose Denominators are Numbers going on in the Progression 1 , 2 , 3 , 4 , &amp;c. unto the Number n which is the Dimensions of the AEquation xnBxn1+Cxn2 &amp;c. ~A= 0 , and whose Numerators are the same Progression inverted .
Divide the second of these Fractions by the first , the third by the second , the fourth by the third , and so on , and place the Fractions which result from this Division above the middle Terms of n 1/ 2 n2~n 2/ 3 ~n1 the AEquation , thus xnBxn1+Cxn2Dxn3 + 3xn3/ 4xn24xn4/ 5xn3 Dx n-3 +Ex n-4 &amp;c. ~A= 0 .
Then if all the Roots of the AEquation are real , the Square of any Coefficient multiplied by the Fraction which stands above , will be greater than the Rectangle under the adjacent Coefficients .
This Corolary doth not hold conversly , for there are an Infinity of AEquations in which the Square of each Coefficient multiplied by the Fraction above it , may be greater than the Rectangle under the adjacent Coefficients , and notwithstanding some or perhaps all of the Roots may be impossible .
Therefore when the Square of a Coefficient multiplied by the Fraction above , is greater than the Rectangle under the adjacent Coefficients , from this Circumstance nothing can be determined as to the Possibility or Impossibility of the Roots of the AEquation : But when the Square of a Coefficient multiplied by the Fraction above it , is less than the Rectangle under the adjacent Coefficients , it is a certain Indication of two impossible Roots .
From what hath been said , is immediately deduced the Demonstration of that Rule which the most illustrious Newton gives for determining the Number of impossible Roots in any given AEquation .
SCHOLIU M. Let the Roots of the AEquation xn Bx n-1 + Cx n-2 Dx n-3 +Ex n-4 Fx n-5 + &amp;c. ~ A=0 ( with their Signs ) be represented by the Letters a , b , c , d , e , f , g , &amp;c. then ( as is commonly known ) B will be the Sum of all the Roots or =a+ b+c+d+e+f+ &amp;c. C the Sum of the Products of all the Pairs of Roots or =ab+ac+ad+ af+ag+ &amp;c. D the Sum of the Products of all the Ternaryes of Roots or =abc+abd+abe+ abf+abg+ &amp;c. E=abed+abce+abcf+ abeg+ &amp;c. F=abcde+abcdf+abcdg+ bedef+ &amp;c. and so on .
Let ( as in this Proposition ) M represent any of these Coefficients , L , N the adjacent Coefficients , and m the Exponent of M ; let Z represent the Sum of the Squares of all the possible Differences between the Terms of the Coefficient M , let ~ be the Sum of all those of the foresaid Squares whose Terms differ by one Letter , ~ the Sum of all those Squares whose Terms differ by two Letters , ~ the Sum of those Squares whose Terms differ by three Letters , ~ the Sum of those Squares whose Terms differ by four Letters and so on .
Thus if M=F=abcde+abcdf+abcdg+ &amp;c. then Z=abcdeabcdf2+abcdeabcdg2+ abcdeabcfg2+bedefabfgh2+ &amp;c. ~=abcdeabcdf2+abcdeabcdg2+ abcdeabcdh2+bedefbcdeg2+ &amp;c. ~=abcdeabcfg2+abcdeabcfh2+ bedefacdfh2+ &amp;c. ~=abcdeabfgh2+ abcdfabegh2+ &amp;c. ~=abcdeafghk2+ acdfgabehk2+ &amp;c. This being laid down I say that the Square of any Coefficient M multiplied by the Fraction m~nm/ m+1~nm+1 exceeds the Rectangle under the adjacent Coefficients L~N/ n+1~Z by n+1~Z/ m+1~nm+1 1/ 2 ~ 1/ 3 ~ 1/ 4 ~ 1/ 5 ~ &amp;c. The Series 1/ 2 ~ 1/ 3 ~ 1/ 4 ~ &amp;c. must consist of m Number of Terms .
Let the AEquation be x5 B x4 +C x3 Dx2+ExA= 0 , whose Roots let be a , b , c , d , e , in which Case n= 5 .
Let M=B=a+ b+c+d+ e , then L= 1 , N= G , m= 1 , Z=ab2+ac2+ad2+ae2+ bc2+ &amp;c. = ~ ; therefore 1~51/ 1+1~51+1~ B2 or 2/ 5 B2 exceeds 1~C by 5+1~Z/ 1+1~51+1 1/ 2 ~= 3/ 5 Z 1/ 2 ~= ( because Z= ~ ) 1/ 10 Z=1/ 10 ab2+1/ 10 ~ac2+1/ 10 ad2+ &amp;c. which is always a positive Number when the Roots a , b , c , d , e are real , positive or negative Numbers .
Let M=C=ab+ac+ad+ ae+bc+ &amp;c. then L= B , N= D , m= 2 , Z=abac2+abad2+abcd2+ abde2+ &amp;c. ~=abac2+abad2+ abae2+ &amp;c. ~=abcd2+abce2+ abde2+ &amp;c. therefore 2~52/ 2+1~52+1~ C2 or 1/ 2 C2 surpasses B~D by 5+1~Z/ 2+1~52+1 1/ 2 ~ 1/ 3 ~= ( because Z=~+ ~ ) = 1/ 6 ~= 1/ 6 ~abcd2+ 1/ 6 abce2+ 1/ 6 ~ abde2+ &amp;c. which is always a positive Number when the Roots a , b , c , d , e are real Numbers , positive or negative .
Let M=D=abc+abd+ abe+acd+ace+ &amp;c. then L= C , N= E , m= 3 , Z=abcabd2+abcabe2+ abcade2+ &amp;c. ~=abcabd2+abcabe2+ abcacd2+ &amp;c. ~=abcade2+abccde2+ abcbde2+ &amp;c. ~= 0 , therefore 3~53/ 3+1~53+1~ D2 or 1/ 2 D2 exceeds C~E by 5+1/ 3+1~53+1~ Z 1/ 2 ~ 1/ 3 ~= ( because Z=~+ ~ ) = 1/ 6 ~ ~ =~1/ 6 ~abcade2+ 1/ 6 ~abccde2+ 1/ 6 abcbde2+ &amp;c. which is a positive Number when the Roots are real Numbers .
Let M=E= abed+abce+abde+ bed &amp;c. then L= D , N= A , m= 4 , Z=abcdabce2+ abcdbcde2+abcdacde2+ &amp;c. = ~ , ~=0=~= ~ , therefore 4~54/ 4+1~54+1~ E~ or 2/ 5 E2 exceeds D~A by 5+1/ 4+1~54+1~Z 1/ 2 ~= 3/ 5 Z 1/ 2 ~= 1/ 10 Z= 1/ 10 ~ abed abce 2+ 1/ 10 ~ abed bed 2+ &amp;c. which is a positive Number when the Roots are real Numbers .
PROPOSITION II .
Let xn Bx n-1 +Cx n-2 Dx n-3 +Ex n-4 &amp;c. ~A=0 be an AEquation of any Degree , whose Roots with their Signs let be expressed by the Letters a , b , c , d , e , f , &amp;c. let M represent any Coefficient of this AEquation , L , N the Coefficients adjacent to M ; K , O the Coefficients adjacent to L , N ; I , P those adjacent to K , O ; H , Q those adjacent to I , P , and so on .
Let m represent the Exponent of M and let Z ( as in the preceeding Proposition ) represent the Sum of the Squares of all the possible Differences between the Terms of the Coefficient M. Then the Product of the Square of any Coefficient M multiplied by the Fraction 1/ 2 ~ 11/ n~n1/ 2~n2/ 3~ &amp;c. ~nm+1/ m doth always exceed L~NK~O+I~PH~Q+ &amp;c. by 1/ 2 Z/ n~n1/ 2~n2/ 3~ &amp;c. ~nm+1/ m which is always a positive Number , when the Roots a , b , c , d , e &amp;c. are real Numbers positive or negative .
Let the AEquation be of the seventh Degree or x7 B x6 +C x5 D x4 +E x3 F x2 +GxA= 0 , whose Roots let be a , b , c , d , e , f , g , in which Case n= 7 .
Let M=E=abed+abce+ abcf+abcg+ bed + &amp;c. then m= 4 , L= D , N= F , K= C , O= G , I= B , P= A , Z=abcdabce2+abcdabcf2+ abcdabcg2+ &amp;c. Therefore 1/ 2 ~ 11/ 7~ 6/ 2 ~ 5/ 3 ~ 4/ 4 ~ E2 or 17/ 35 E2 exceeds D~ FC~G+B~A by 1/ 2 Z/ 7~ 6/ 2 ~ 5/ 3 ~ 4/ 4 or Z/ 70 = 1/ 70 ~abcdabce2+ 1/ 70 ~abcdabcf2+ &amp;c. From this Proposition , is deduced the following , Rule for determining the Number of impossible Roots in any given AEquation .
From each of the Unciae of the middle Terms of that Power of a Binomial , whose Index is the Dimensions of the proposed AEquation , subtract Unity , then divide each Remainder by twice the Correspondent Uncia , and set the Fractions which result from this Division , above the middle Terms of the given AEquation .
And under any of the middle Terms if its Square multiplied by the Fraction standing above it , be greater than the Rectangle under the immediately adjacent Terms , Minus the Rectangle under the next adjacent Terms , Plus the Rectangle under the Terms then next adjacent &amp;c. place the Sign + , but if it be less , place the Sign - .
And under the first and last Term place + .
And there will be at least as many impossible Roots , as there are Changes in the Series of the under-written Signs from + to - , or from to + .
Let it be required to determine the Number of impossible Roots in the AEquation x7 5 x6 +15 x5 23 x4 +18 x3 +10 x2 28x+24= 0 .
The Unciae of the middle Terms of the 7th Power of a Binomial are 7 , 21 , 35 , 35 , 21 , 7 , from which subtracting Unity , and dividing each of the Remainders by twice the correspondent Uncia , the Quotients will be 6/ 14 , 20/ 42 , 34/ 70 , 34/ 70 , 20/ 42 , 6/ 14 or 3/ 7 , 10/ 21 , 17/ 35 , 17/ 35 , 10/ 21 , 3/ 7 which Fractions place above the middle 3/ 7 10/ 21 Terms of the AEquation , has x5 x6 +15 x5 ++ 17/ 35 17/ 35 10/ 21 3/ 7 23 x4 +18 x3 +10 x2 28x+24= 0 .
Then +++ because the Square of 5 x6 multiplied into the Fraction over its Head 3/ 7 , to wit 75/ 7 x12 is less than x7 ~15 x5 or 15 x12 I place the Sign under the Term 5 x6 .
Because the Square of 15 x5 multiplied by the Fraction over its Head 10/ 21 to wit 705/ 7 x10 is greater than ~5 x6 ~ 23 x4 ~ x7 ~18 x3 =97 x10 I place the Sign + under the Term 15 x5 .
Seeing 8993/ 35 x8 ( the Square of the Term 23 x4 multiplied by the Fraction over its Head 17/ 35 ) is less than 15 x5 ~18 x3 ~5 x6 ~10 x2 + x7 ~28x=292 x8 , I place the Sign under the Term 23 x4 .
Because 18 x3 2~ 17/ 35 or 5508/ 35 x6 exceeds ~2 3 x4 ~ 10x2 ~1 5 x5 ~28x+ ~5 x6 ~24=70 x6 I place the Sign + under the Term 18 x3 .
Since 10 x2 2~ 10/ 21 or 1000/ 21 x4 is less than ~+ 18 x3 ~28x ~2 3 x4 ~24=48 x4 I place the Sign under the Term 10 x2 .
Because 28 x2 ~ 3/ 7 or 336 x2 is greater than 10 x2 ~24=240 x2 under 28x I place + , then under the first and last Terms I place + ; and the six Changes of under-written Signs shows that there are six impossible Roots .
If the impossible Roots were to be found by the Newtonian Rule , the Operation would stand thus : 3/ 7 5/ 9 3/ 4 3/ 3 5/ 9 x7 5 x6 +15 x5 23 x4 +18 x3 +10 x2 +++++ 3/ 7 28x+24= 0 , by which Rule there are found ++ only two impossible Roots , whereas there are six to wit 1+~ 3 , 1~ 3 , 1+~ 2 , 1~ 2 , 1+~ 1 , 1+~ 1 , the seventh Root being 1 .

