arXiv:0707.0699v1 [math.

HO] 4 Jul 2007

A double demonstration of a theorem of Newton, which gives a relation between the coecient of an algebraic equation and the sums of the powers of its roots
Leonhard Euler

1. After an algebraic equation has been given, either from fractions or by irrationals, and reduced to this form xn Axn1 + Bxn2 Cxn3 + Dxn4 Exn5 + N = 0, one demonstrates in analysis that the equation always has as many roots, either real or imaginary, as unities are contained in the exponent n of the highest power. Then indeed it is no less certain that if all the roots of this equation are , , , , , . . . , , the coecients of the terms of the equation A, B, C, D, E etc. are comprised thus from them A B C D E = = = = = sum of sum of sum of sum of sum of etc., all the roots the products the products the products the products of of of of two three four ve = = = = = + + + + + , + + + + etc., + etc., + etc., + etc.

and nally the last term N is the product of all the roots . 2. Now it will help me to simply and briey explain this theorem, whose demonstration I give in this paper, if I rst introduce some notation. Let designate the sum of all the roots, 2 the sum of the squares of the roots, 3 the sum of the cubes of the roots, 4 the sum of the biquadrates of the roots,
Presented to the Berlin Academy on January 12, 1747. Originally published as Demonstratio gemina theorematis Neutoniani, quo traditur relatio inter coecientes cuiusvis aequationis algebraicae et summas potestatum radicum eiusdem, Opuscula varii argumenti 2 (1750), 108120. E153 in the Enestr om index. Translated from the Latin by Jordan Bell, School of Mathematics and Statistics, Carleton University, Ottawa, Canada. Email: jbell3@connect.carleton.ca

and so on, so that it will be 2 3 4 5 = = = = = 2 3 4 5 etc. + + 2 + 3 + 4 + 5 + + 2 + 3 + 4 + 5 + + 2 + 3 + 4 + 5 + + 2 + 3 + 4 + 5 + + + + + +, + 2 , + 3 , + 4 , + 5

3. With the rule for this sign explained, Newton arms that these sums of powers, which are formed from all the roots, can thus be dened by the coecients of the equation A, B, C, D, E etc.

2 3 4 5 6

= A, = A = A = A = A = A etc. 2B, 2 B 3 B 4 B 5 B + 3C, 2 + C 3 + C 4 + C 4D, 2 D 3 D + 5E, 2 + E 6F

Newton does not give any demonstration of this theorem, but rather seems to conclude its truth by continued examples. First, as no demonstration is required that = A, and since A2 = 2 + 2 + 2 + 2 + 2 + etc. + 2 + 2 + 2 + 2 + 2 + etc., it will be A2 = and hence 2 = A2 2B = A 2B ; in a similar way the truth of the following formulae can be shown, but continuing will take more work. 4. Though many have already shown the truth of this most useful theorem, their demonstrations are often based on rules of combinations, which, though 2 2 + 2B

true, depend however on induction. I will thus oer here a double demonstration, the rst of which completely avoids induction. Indeed the one from the analysis of the innite is small, and though it may seem quite remote, everything comes together perfectly; however since it can reasonably be objected that the truth of this theorem should be obtained before the analysis of the innite is come to, I add another demonstration, in which nothing is assumed beyond that which follows immediately from the nature of equations.

Demonstration 1.
5. Let us put xn Axn1 + Bxn2 Cxn3 + N = Z, and if the roots of the equation Z = 0, or values of x, are , , , , . . . , , whose number is = n, it will be from the nature of equations Z = (x )(x )(x )(x ) (x ) and by taking the logarithm we will obtain lZ = l(x ) + l(x ) + l(x ) + l(x ) + + l(x ). But if the dierential of these formulae are now taken, it will be dZ dx dx dx dx dx = + + + + + Z x x x x x and so by dividing by dx it will become dZ 1 1 1 1 1 = + + + + + . Zdx x x x x x Now let each of these fractions be converted into innite geometric series in the usual way, as 1 x 1 x 1 x 1 x = = = 1 + x 1 + x 1 + x etc. 1 + x 2 3 4 5 + + + + + etc., x2 x3 x4 x5 x6 2 3 4 5 + + + + + etc., x2 x3 x4 x5 x6 2 3 4 5 + + + + + etc. x2 x3 x4 x5 x6 2 3 4 5 + + + + + etc. x2 x3 x4 x5 x6 3

Therefore by collecting these series together, using the signs explained above , 2 , 3 etc., because the number of these series is = n, dZ n 1 = + 2 Zdx x x + 1 x3 2 + 1 x4 3 + 1 x5 4 + etc.

6. On the other hand, since we have put Z = xn Axn1 + Bxn2 Cxn3 + Dxn4 N, it will similarly be, by taking dierentials, dZ = nxn1 (n 1)Axn2 +(n 2)Bxn3 (n 3)Cxn4 +(n 4)Dxn5 etc. dx and then let the above formula be put in the same form as be
dZ Zdx ,

so that it would

dZ nxn1 (n 1)Axn2 + (n 2)Bxn3 (n 3)Cxn4 + (n 4)Dxn5 etc. = , Zdx xn Axn1 + Bxn2 Cxn3 + Dxn4 etc.

and this fraction will therefore be equal to the series found previously 1 n + 2 x x
n

1 x3

2 +
dZ Zdx n2

1 x4

3 +

1 x5

4 + etc.

Now if each expression found for x Ax this equation will result

n1

is multiplied by the one denominator Cxn3 + Dxn4 etc.,

+ Bx

nxn1 (n 1)Axn2 + (n 2)Bxn3 (n 3)Cxn4 + (n 4)Dxn5 etc. = nxn1 +xn2 +xn3 2 +xn4 3 nAxn2 Axn3 Axn4 2 +nBxn3 +Bxn4 nCxn4 +xn5 4 Axn5 3 +Bxn5 2 Cxn5 +nDxn5 +etc. etc. +etc. etc. +etc.

7. Now just as much as both the rst terms nxn1 are equal, it is also necessary that the second, third, fourth etc. are equal to each other; thus the following equations arise (n 1)A = +(n 2)B (n 3)C +(n 4)D = = = nA, 2 A 3 A 4 A etc. 4 + nB, 2 + B 3 + B nC, 2 C + nD

and the law by which these equations progress is at immediately clear. Further, from these, the formulae can be obtained which comprise the theorem of Newton, namely 2 3 4 5 = A, = A = A = A = A etc. This is the rst demonstration of the proposed theorem. 2B, 2 B 3 B 4 B + 3C, 2 + C 3 + C 4D, 2 D + 5E

Demonstration 2.
8. So this demonstration can be clearly understood, I will x an equation of a determinate degree, but one can see that this extends equally well to any degree. Thus let the given equation be of the fth degree x5 Ax4 + Bx3 Cx2 E = 0, all of whose roots are , , , , . Then because any root substituted in place of x satises the equation, it will be 5 5 5 5 5 A4 A 4 A 4 A 4 A4 +B3 +B 3 +B 3 +B 3 +B3 C2 C 2 C 2 C 2 C2 +D +D +D +D +D E E E E E = 0, = 0, = 0, = 0, = 0.

These equations can be collected into one sum, and using the above sign (2) one will have 5 A or 5 = A 4 B 3 + C 2 D + 5E. 4 + B 3 C 2 + D 5E = 0

9. It is clear then that if the given equation were of any degree xn Axn1 + Bxn2 Cxn3 + Dxn4 M x N = 0, 5

where in the last term, the top of the ambiguous signs occurs if the exponent n is an odd number, and the bottom if even, it will likewise be n = A n1 B n2 + C n3 M nN,

where indicates any root of this equation. Thus the truth of the Theorem of Newton is now shown for this one case. It therefore remains for us to demonstrate the truth of it for both the higher and for the lower powers of the roots. 10. Indeed for the higher powers the same argument applies; for if the values , , , , satisfy the equation x5 Ax4 + Bx3 Cx2 + Dx E = 0, then they also will satisfy the following equations x6 Ax5 + Bx4 Cx3 + Dx2 Ex x7 Ax6 + Bx5 Cx4 + Dx3 Ex2 = = 0, 0, 0,

x8 Ax7 + Bx6 Cx5 + Dx4 Ex3 = etc.

Therefore also if in each equation all the values , , , , are substituted for x and the aggregate collected, it will be 6 = A 7 = A 8 = A etc. 11. Thus if denotes any root of this equation xn Axn1 + Bxn2 Cxn3 + Dxn4 M x N = 0, it will not only be, as we have already found, n = A n1 B n2 + C n3 D n4 + M nN, 5 B 6 B 7 + C 4 + C 5 + C 5 D 3 D 4 D 4 + E 2 + E 3 + E 3 , 2 ,

but by proceeding to higher powers it will also be

Z Z Z n+1 = A n+2 n+3 Z Z Z Z n1 + C n2 D n3 + M 2 N , Z Z Z Z Z Z = A n+1 B n + C n1 D n2 + M 3 N 2 , Z Z Z Z Z Z = A n+2 B n+1 + C n D n1 + M 4 N 3 , n B Z Z

etc.

and indeed in general, if any number m is added to n, it will be

Z n+m = A Z n+m1 B Z n+m2 + C Z n+m3 M Z m+1 N Z m .

It should be noted here that if m = 0, because all the powers 0 = 1, 0 = 1, 0 = 1 etc. and the number of these letters = n, then 0 = n, and in this case the rst formula that was found is included in this expression. 12. This expression could also be used if a negative number is taken for m, and then for the assumed equation of the fth degree x5 Ax4 + Bx3 Cx2 + Dx E = 0 the following formulae would respectively occur 4 3 2 = = = A A A etc., but these formulae are dierent from those which the theorem contains. For it needs to be demonstrated that 4 3 2 = A = A = A = A. 3 B 2 B 2B, 2 + C + 3C, 4D, 3 B 2 B 1 B 2 + C 1 + C 0 + C 1 D 0 D 1 D 0 + E 1 + E 2 + E 1 , 2 , 3

I will therefore show the truth of these formulae in the following way. 13. Namely, given an equation of the fth degree x5 Ax4 + Bx3 Cx2 + Dx E = 0 let us form the following equations of lower degrees by using its coecients I. II. III. IV. x A = 0. x2 Ax + B = 0. x3 Ax2 + Bx C = 0. x4 Ax3 + Bx2 Cx + D = 0. Let Let Let Let p q r s be be be be its root. any root. any root. any root.

The roots of these equations, even though they are dierent from each other, constitute however for all these equations the same sum = A. Next, with the rst removed, the sum of the products from two roots will everywhere be the same = B . Then the sum of the products from three roots will everywhere = C , besides of course equations I and II, where C does not occur. Similarly in the given equation IV etc. the sum of the products from four roots will be the same = D. 14. Therefore in the equations of lower degrees, whose roots are denoted respectively by the letters p, q, r, s, if the letter denotes an arbitrary root of the given equation, it will be = 2 = 3 = 4 = s= s2 = s3 = s4 . r= r2 = r3 , q= q2 , p,

But on the other hand, it follows, from what we demonstrated above in 9, that p q2 r3 s4 = A, = A = A = A q 2B, r2 B s3 B r + 3C, s2 + C s 4D.

Now therefore for the given equation of the fth degree x5 Ax4 + Bx3 Cx2 + Dx E = 0 these formulae arise 2 3 4 = A, = A = A = A 2B, 2 B 3 B + 3C, 2 + C 4D.

16. Therefore in a given equation of any degree xn Axn1 + Bxn2 Cxn3 + Dxn4 N = 0 if the letter indicates an arbitrary root, it will be 2 3 4 5 = A, = A = A = A = A etc. In this way the truth of the Theorem of Newton is also demonstrated. 2B, 2 B 3 B 4 B + 3C, 2 + C 3 + C 4D, 2 D + 5E