The Geometry of the Complex Plane and the Zeros of Polynomials
[56/100%]Let be arbitrary points lying in the finite part of the complex plane and let be non-negative masses whose sum is which are placed at the points respectively. In any linear transformation of the complex plane that carries the point at infinity into itself, the center of gravity of such a mass-distribution is mapped by the same transformation, i.e. if we again place the point mass at the image-point of (), then the center of gravity of this new mass distribution is identical with the image of the point .
In the sequel (102--156) we mean by a point'' of the complex plane an arbitrary finite point or the point at infinity, by a circle'' the circumference of a circle or a straight line---the latter passes through the point at infinity---and by a ``circular domain'' a domain whose boundary is a circle. A circular domain is either the closed interior or the closed exterior of a circle or a closed half-plane, according as it does not contain the point at infinity or contains it in its interior or on its boundary.
Let the points and the masses be given. Let be different from all the points while of the latter some may coincide. Let .
Find a point situated in such a manner that, in a linear transformation of the plane that maps the points
into the points
respectively, becomes the center of gravity of the mass-distribution determined by the masses placed at the points ().
There are infinitely many linear transformations that map into ; the point however is independent of the choice of the underlying transformation.
The point , defined above, which is determined uniquely by the system of points , by the masses , , placed there, and by the ``point of reference'' which must be distinct from all the points , is called the center of gravity of the mass distribution with respect to . If is the point at infinity, , then is the ordinary center of gravity.
Consider all possible mass distributions with total mass over the fixed points and let be a point distinct from . The centers of gravity of all mass distributions of this kind with respect to cover a ``circular-arc polygon'' .
Let be the circular-arc polygon defined in 103 and two points of . Consider the circle through and . The arc of this circle with end-points , that does not contain , is contained in , in other words is circularly-convex with respect to . (We call the smallest circularly-convex domain with respect to that contains the points .)
If the points all belong to a circular domain and if the point lies outside , then the circular-arc polygon [103] lies completely inside .
Let be arbitrary points and a point distinct from these. Let the same mass be placed at each of the points . In the sequel we consider only the center of gravity of this special mass distribution and call it for brevity the center of gravity of the points with respect to .
Let be the center of gravity of with respect to . Every circle through and separates the points , i.e. either both circular domains bounded by the circle contain some of the points in their interior or all the points lie on the circle.
If all the points belong to a circular domain , then the points and cannot both lie outside . If one of them, e.g. , lies outside , then the other, , lies in the interior of , except in the case where all the points coincide at a boundary point of and then also lies there.
Let the fixed points be pair-wise distinct. Let the complex numbers take only the different values . The centers of gravity with respect to of all systems of this kind lie everywhere dense in the smallest domain, circularly convex with respect to , that contains the points .
The points are fixed and the variable point converges to . To what limiting position does the center of gravity of the points with respect to tend?
The center of gravity of the finite fixed points with respect to the variable point may for sufficiently large be expanded in decreasing powers of . Calculate the first two terms of the expansion.
In the sequel (111--156) we shall mean by a rational integral function (polynomial) of degree a function not vanishing identically of the form
Here is not necessarily different from . In the special case when , we say that is the precise degree of . If , , we shall say that has the -fold zero . With this interpretation every rational integral function of degree has a total of zeros, of which some may possibly lie at infinity. Thus there is determined by a system of points , in the complex plane, the zeros of . Conversely, to every system of points there belongs a polynomial, whose coefficients are uniquely determined up to a constant of proportion. We shall often disregard this unimportant constant factor and speak of the polynomial, whose zeros are .
By the center of gravity of a polynomial we mean the center of gravity of its zeros. Express the center of gravity of with respect to a variable point : (1) in terms of and , (2) in terms of the coefficients of . Consider the special cases and .
Let be the center of gravity of with respect to . A necessary and sufficient condition that has only real zeros is that the imaginary parts of and have opposite signs if they do not both vanish however may vary. ( counts as a real zero.)
With respect to which points of the plane does the center of gravity of lie at infinity? What do the statements in 106 and 107 say about these points?
Let be a polynomial of precise degree , a circular domain that contains the zeros of and . The zeros of the derivative of the transcendental function lie either in or in , i.e. in the circular domain that is obtained from by parallel displacement by the vector .
Cf. Problem 33 of Part Three (Gauss's theorem in the circular-domain formulation used in 113); consider the center of gravity of with respect to one of the zeros in question.
Let be a zero of the polynomial of degree , finite and , and . Then has at least one zero in every circle that passes through both the points
Let be a polynomial of degree , whose zeros are all of absolute value . Let and be arbitrary positive numbers, such that . Then all the zeros of are in absolute value
Let be a polynomial of degree , all of whose zeros lie in the annulus . Let and be arbitrarily positive numbers such that . Then all the zeros of lie in the annulus
If is a finite zero of the polynomial of degree , then the center of gravity with respect to of the remaining zeros of is
How is this formula to be modified if is at infinity?
The th Hermite polynomial satisfies the differential equation
(This is the differential equation satisfied by the Hermite polynomial as defined via in problem 100(g) of Part Six's solution.) Show by means of 118 that the Hermite polynomials have only real zeros (problem 100(i) of Part Six's solution). [Consider the assumed complex zero with largest imaginary part.]
The polynomial of degree that satisfies the differential equation
(the th Legendre polynomial, cf. problem 90 of Part Six), has only real zeros (problem 97 of Part Six).
Gauss's theorem (the circular-domain formulation given in the introduction to problem 31 of Part Three) is equivalent to the following: If all the zeros of a polynomial lie in a circular disc , then the zeros of its derivative also lie in [113]. Decide whether the following theorem is true or false: If all the zeros of a polynomial lie in two circular discs , then the zeros of its derivative also lie in or .
Let and be two circular discs or, more generally, two arbitrary circular domains. Denoting by and fixed positive numbers, we call the set of all points
where and independently run through and respectively, a mean domain of and , and use for it the following notation
is completely determined by and . [Cf. H. Minkowski: Werke, Vol. 2, p. 176, Leipzig: B. G. Teubner 1911.]
Let and be two circular discs whose centers are and radii respectively. Show that then
is again a circular disc whose center and radius are given by the following equations:
is homothetic to and with respect to the same center of similitude. [H. S. M. Coxeter: Introduction to Geometry. New York: John Wiley Sons 1961, pp. 68--70.]
Let and be two half-planes with parallel boundary lines, one of which contains the other. The mean domain is then again a half-plane, whose boundary line runs parallel to those of and and divides the distance between them in the ratio .
Let , and be a polynomial of degree , a polynomial of degree . Let all the zeros of lie in the circular domain and all the zeros of in the circular domain . Then the zeros of the derivative lie either in or in or in the mean domain
(By degree we mean the precise degree.)
Let be a rational (non-integral) function, whose numerator and denominator have no common factors. Let it have zeros (poles) which lie in the circular domain and poles (zeros) which lie in the circular domain . (Only finite zeros and poles are considered.) Let . Then the zeros of lie either in or in or in the domain defined by the numbers
where and independently run through and , for which we use the notation
If and the circular domains and have no common points, then all the zeros of lie either in or in .
Let be a polynomial, and let the equation
have all its roots in a convex domain (oval) . Then all the numbers for which this is the case also lie in a convex domain .
Let be a polynomial and let its -points (the zeros of ) lie in a circular domain and its -points in a circular domain . If are positive numbers and we set , those -points of that lie neither in nor in , lie in the mean domain
[We may take to be integers.]
By the derivative of a polynomial of degree with respect to the point , denoted by , we mean the ordinary derivative if , and
if is finite. is a polynomial of degree . (When using homogeneous coordinates this expression is called the first polar, or, by older authors, emanant; cf. Laguerre, Oeuvres, Vol. 1, p. 48. Paris: Gauthier-Villars, 1898.)
Let the coefficients of (in the notation of §2 preceding 111) be . What are the coefficients of ?
Let be arbitrary. Show that the operation is distributive: If we denote by and two polynomials of degree , and by and arbitrary constants, then we have
Let be arbitrary and a polynomial of degree , a factorization of into the two factors and of degree and respectively, . Show that
Let be a polynomial of degree , and two arbitrary constants. The operations and are commutative, that is to say
Show that vanishes identically if and only if all the zeros of coincide with .
By the derived system of the points with respect to an th point we mean the zeros of , where is the polynomial of degree whose zeros are , all the zeros being correctly counted (§2, preceding 111). The derived system is completely determined, except in the single exceptional case where the points all coincide (cf. 132).
If is finite, then to the derived system with respect to in general there belong four different kinds of points:
(1) The finite points distinct from , with respect to which the center of gravity of is precisely .
(2) The multiple zeros of , distinct from , each with multiplicity reduced by .
(3) itself only if it is a zero of and in that case with the same multiplicity.
(4) The point if it is at least a double zero of ; moreover also if it is not a zero of but if is the ordinary center of gravity of , i.e. the one constructed with respect to .
A circular domain, whose boundary contains as well as a point of the derived system with respect to , itself contains points of the original system.
If a circular domain contains the points and does not contain the point , then it contains also the derived system of with respect to .
The derived system of the points lies in the smallest circular-arc polygon, circularly convex with respect to , that contains . (Generalization of Problem 31 of Part Three.)
If is of degree , then is of the first degree. Calculate the only zero of !
Let
be an arbitrary polynomial of degree and let be arbitrary points in the complex plane. The expression
is a symmetric multilinear function of the variables , which we can write as follows:
Here are the elementary symmetric functions of :
If the last of the numbers (and only these) lie at infinity, then we set , , , .
Let
be an arbitrary polynomial of degree with zeros and
an arbitrary polynomial of degree with zeros . Calculate and .
If the two polynomials are related in such a manner that
or, equivalently [solution 139], , then the two polynomials and are termed apolar. The term suggests the vanishing of the th polar . The condition of apolarity is
[139.] The two systems and are also termed mutually apolar.
What is the geometrical significance of the apolarity of , and that of and ?
Which are the systems , that are apolar to the system consisting of the roots of the equation ?
Let be an arbitrary system of points. There are systems of coincident points that are apolar to the given system. Which are they?
The polynomial of degree
is apolar to any system of numbers whose sum is and product is .
Let be an arbitrary polynomial of degree , all of whose zeros lie in the circular domain . If , , are arbitrary points lying outside , then all the zeros of the polynomial of degree , lie in .
Let the two polynomials and of degree be apolar to each other. Every circular domain that contains all the zeros of one polynomial also contains at least one zero of the other.
Let the two polynomials and of degree be apolar to each other. The two smallest convex polygons that contain all the zeros of and respectively have at least one common point. More generally, the two smallest circular-arc polygons that are circularly convex with respect to the same point and that contain all the zeros of and respectively, also have at least one common point.
The polynomial
always has one zero in the circle .
The polynomial
always has one zero in the circle .
The polynomial of terms
always has one zero in the circle
and hence in the circle
(Generalization of 147.)
If a polynomial of degree assumes the same value at two points and , , then its derivative has at least one zero in the circular disc with center at the midpoint of the line segment and radius . (Analogue of Rolle's theorem in the complex domain.)
Let
be a polynomial of degree all of whose zeros lie in a circular domain . Also let
be a polynomial of degree with zeros . Every zero of the polynomial
formed by ``composition'' of and then has the form
where denotes a suitably chosen index and a suitably chosen point in . (Here we must set and indeterminate.)
Let the zeros of the polynomials and of degree all lie in the unit circle , and the zeros of at least one of the two polynomials in fact in . Then also all the zeros of the polynomial of degree formed from and by composition [151] lie in the interior of the unit circle .
Let the polynomial of degree have all its zeros in a convex domain that contains the origin. Let the polynomial of the same degree have only real zeros which lie in the interval . The polynomial of degree derived from the two polynomials by composition [151] then also has all its zeros in .
If all the zeros of the polynomial
of degree lie in the real interval , and those of the polynomial
of degree in the interval (or ) then the zeros of the polynomial
of degree lie in the interval . ( are positive numbers.)
Let all the zeros of the polynomial
of degree be real and those of the polynomial
of degree be real and of the same sign. Then all the zeros of the polynomial
are also real. ( counts as a real zero.)
Under the assumptions of 155, all the zeros of the polynomial
are real.