Escolar Documentos
Profissional Documentos
Cultura Documentos
a r t i c l e i n f o a b s t r a c t
Article history: A univariate polynomial f (x) is said to be nice if all of its coef-
Received 10 May 2014 ficients as well as all of the roots of both f (x) and its derivative
Received in revised form 12 January f (x) are integers. The known examples of nice polynomials
2015
with distinct roots are limited to quadratic polynomials, cu-
Accepted 22 January 2015
Available online 5 March 2015 bic polynomials, symmetric quartic polynomials and, up to
Communicated by David Goss equivalence, only a finite number of nonsymmetric quartic
polynomials and one quintic polynomial. In this paper we find
MSC: parametrized families of nice nonsymmetric quartic polynomi-
11C08 als with distinct roots, as well as infinitely many nice quintic
11D25 and sextic polynomials with distinct roots.
© 2015 Elsevier Inc. All rights reserved.
Keywords:
Nice polynomial
Quartic polynomial
Quintic polynomial
Sextic polynomial
1. Introduction
A univariate polynomial f (x) is said to be nice if all of its coefficients as well as all
of the roots of both f (x) and its derivative f (x) are integers. While nice polynomials
of any arbitrary degree are known, they have repeated roots, a simple example being
xn−1 (x − n). This paper is concerned with nice polynomials that have distinct roots.
It is easy to observe that if f (x) is a nice polynomial of degree d and a, b, c are
arbitrary integers such that a = 0, c = 0, then g(x) = ad cf ((x + b)/a) is also a nice
http://dx.doi.org/10.1016/j.jnt.2015.01.010
0022-314X/© 2015 Elsevier Inc. All rights reserved.
A. Choudhry / Journal of Number Theory 153 (2015) 354–363 355
polynomial of degree d. This observation seems to have been made first of all by Caldwell
[3]. In fact, the polynomial g(x) is at times a nice polynomial even for certain noninteger
rational values of a, b, c. Thus, given a single nice polynomial f (x) of degree d, we
readily obtain infinitely many nice polynomials of degree d. Two nice polynomials that
are obtained in this manner from the same nice polynomial will be considered equivalent.
More precisely, two nice polynomials f1 (x) and f2 (x) of degree d will be considered
equivalent if there exist rational numbers r, s, t such that rs = 0 and f1 (x) = rf2 (sx +t).
When we refer to distinct nice polynomials, we will mean nice polynomials that are not
equivalent.
A complete determination of all nice quadratic and cubic polynomials with distinct
roots has already been done [1,2,7]. Nice quartic polynomials with distinct roots were first
found by Caldwell [3] but this paper has remained inaccessible to me. It appears from the
detailed discussion in [6, pp. 21–23] that while all nice symmetric quartic polynomials
have been found, only finitely many distinct nice nonsymmetric quartic polynomials
are known. Further, up to equivalence, only one numerical example of a nice quintic
polynomial has been found [6, p. 23]. No nice sextic polynomials have been found till
now.
In this paper we find nice nonsymmetric quartic polynomials whose coefficients are
given in parametric terms and whose roots are all distinct. This yields infinitely many
distinct nice nonsymmetric quartic polynomials. We also find infinitely many distinct
nice quintic and sextic polynomials whose roots are all distinct.
It is readily seen that if f (x) is an nth degree polynomial with n rational roots such
that all of the n − 1 roots of f (x) are also rational, then for a suitably chosen integer
value of h, the polynomial hn f (X/h) is such that its coefficients are integers and all of the
roots of both hn f (X/h) and its derivative are integers. Thus to obtain nice polynomials,
it suffices to construct polynomials f (x) such that all the roots of both f (x) and its
derivative f (x) are rational.
Since we want the derivative f (x) to have three distinct rational roots, we must have
where p, q, r are distinct rational numbers. Comparing the coefficients of the two values
of f (x) given by (2) and (3), we get the following diophantine system:
The complete solution of Eq. (6) in nonzero rational numbers may be written as
where u and v are arbitrary nonzero rational parameters. Substituting these values of
a, b, c, in Eq. (4), and solving for r, we get,
On substituting the values of a, b, c and r in Eq. (5), we get the following quadratic
equation in p and q:
One solution of this equation is obtained quite simply by taking v = 4/3, when the
coefficient of q 2 in (9) vanishes, and we get a nonzero solution for p, q which, on working
backwards, yields the following solution for the system of equations (4), (5) and (6):
f (x) = (x + 2u4 + 12u3 − 55u2 + 36u + 18)(x − 6u4 + 30u3 − 31u2 − 18u + 18)
× (x + 2u4 − 6u3 − 31u2 + 90u − 54)(x + 2u4 − 36u3 + 117u2 − 108u + 18),
= x4 + φ1 (u)x2 + φ2 (u)x + φ3 (u),
f (x) = 4(x + 2u4 − 24u3 + 74u2 − 72u + 18)(x − 4u4 + 20u3 − 37u2 + 12u + 18)
× (x + 2u4 + 4u3 − 37u2 + 60u − 36).
The nice polynomials obtained above will have a repeated root if the following condi-
tion is satisfied:
where the values of a, b, c are given by (10) in terms of the parameter u. The condition
(13) can be satisfied only by finitely many values of the parameter u, and excluding
these values, we get nice quartic polynomials with distinct roots. Further, it is readily
established that any polynomial of type (1) is symmetric if and only if the following
condition is satisfied:
As in the case of condition (13), the condition (14) is satisfied only by finitely many
values of the parameter u, and by excluding these values as well, we ensure that we
obtain infinitely many nice nonsymmetric quartic polynomials with distinct roots.
As an example, the discriminant (12) becomes a perfect square when we take v =
2(2u + 3)(2u − 3)2 /{(11u2 − 30u + 21)u}, and this leads to the following solution of the
system of Eqs. (4), (5) and (6):
where a, b, c, d are distinct rational numbers. Equating f (x) to 5(x − p)(x − q)(x −
r)(x − s), and comparing coefficients, we get the following diophantine equations,
4a + 4b + 4c + 4d = 5p + 5q + 5r + 5s, (17)
3ab + 3ac + 3ad + 3bc + 3bd + 3cd = 5pq + 5pr + 5ps + 5qr + 5qs + 5rs, (18)
2abc + 2abd + 2acd + 2bcd = 5pqr + 5pqs + 5prs + 5qrs, (19)
abcd = 5pqrs. (20)
and so it follows from (21) and (22) that 2abc = 5pqr and 2abd = 5pqs. Thus Eq. (19)
reduces to 2acd + 2bcd = 5prs + 5qrs, and using (21), it further reduces to the following
linear equation:
8a + 8b = 5p + 5q. (23)
We solve Eqs. (17) and (23) for b and q and use the values of c, d given by (21) to get,
where u is some rational number. Substituting the values of c, d given by (21) and the
values of b, q given by (24) in (18), and using the values of a, s given by (26), Eq. (18)
reduces to the following equation,
(32u4 + 24u3 − 143u2 + 30u + 50)p2 − 28(u − 2)(4u2 − 5)pr + 84(u − 2)2 r2 = 0. (27)
Now (27) is a quadratic equation in p and r, and it will have a rational solution if its
discriminant 16(u − 2)2 (112u4 − 504u3 + 1043u2 − 630u + 175) is a perfect square. This
leads to the diophantine equation,
360 A. Choudhry / Journal of Number Theory 153 (2015) 354–363
and
reduces (28) to the Weierstrass minimal form of the elliptic curve given by
It was determined using APECS that (31) is an elliptic curve of rank 2 and the Mordell–
Weil basis is given by the two rational points (−92, 4320) and (−30, 3430). The rank of
the curve (31) is also readily confirmed by a reference to Cremona’s database on elliptic
curves [4]. We can now find infinitely many rational points on the curve (31), and thus
obtain infinitely many corresponding rational points on the curve (28). Working back-
wards, we find infinitely many solutions of the simultaneous diophantine equations (17),
(18), (19) and (20), and thus obtain infinitely many nice quintic polynomials.
The polynomials thus obtained will have a repeated root if the following condition is
satisfied:
Using the values of a, b, c, d, given by (21), (24) and (26), we find that both (27) and
(32) can be satisfied only by a finite number of values of u. Since we can obtain infinitely
many values of u that satisfy (28), we can simply ignore the finitely many values of u
that yield nice polynomials with repeated roots, and thus obtain infinitely many nice
quintic polynomials with distinct roots.
We give two numerical examples corresponding to the two rational points (−92, 4320)
and (−30, 3430) lying on the curve (31). The first nice quintic polynomial thus obtained
and its derivative are given by
while the second nice polynomial and its derivative are as follows:
The resultant obtained by eliminating p between these two equations may, after suitable
transpositions, be written as follows:
b2 + c2 − 2q 2 = m(b2 − c2 ), (39)
m(4a2 + b2 + c2 − 6q 2 ) = b2 − c2 , (40)
Substituting these values of b, c, q, in Eqs. (36) and (40), we get two equations which
we simplify further by taking g = 1, h = −1, and we thus obtain the following two
equations,
Now Eq. (43) has the solution (a, m) = (2, 1), and using this known solution, we get
the following solution of (43):
Substituting these values of a and m in Eq. (42), and writing p = 2s/{3(t2 +8)}, Eq. (42)
reduces to the following quartic equation,
Now Eq. (45) is a quartic model of an elliptic curve, and the birational transformation
given by
and
Using APECS, it was determined that (48) is an elliptic curve of rank 1, and the Mordell–
Weil basis is the rational point P whose coordinates are given by (X, Y ) = (14, 200).
The rank is also confirmed by a reference to Cremona’s database of elliptic curves [4].
We can thus determine infinitely many rational points on the curve (48), and working
backwards obtain infinitely many solutions of the diophantine equations (36) and (37)
leading to infinitely many nice symmetric sextic polynomials. It is easily established that
there are only a finite number of values of t that lead to nice sextic polynomials with
repeated roots. Excluding such values of t, we can obtain infinitely many nice sextic
polynomials with distinct roots.
We note that while the point P on the elliptic curve (48) leads to the nice sextic
polynomial (x2 −100)2 (x2 −4) which has repeated roots, the point 2P yields the following
nice sextic polynomial,
A second example of a nice sextic polynomial with distinct roots, obtained from the
rational point 3P on the curve (48), is given by
It would be of interest to find nice polynomials of degree greater than 6 with all
distinct roots. No such examples are known.
It is also an open problem whether there are any nice quartic polynomials f (x) with
distinct roots such that the roots of the second derivative f (x) are also integers. Limited
computer trials on the two families of nice polynomials obtained in Section 2 did not yield
any such quartic polynomial. It has been conjectured in [1] that such quartic polynomials
do not exist. It is, however, difficult to rule out the existence of such quartic polynomials
in view of the infinitely many families of nice quartic polynomials with distinct roots
that have been found in this paper.
Acknowledgments
I am grateful to the referee for his comments which have led to improvements in the
paper.
References
[1] R.H. Buchholz, J.A. MacDougall, When Newton met Diophantus: a study of rational-derived poly-
nomials and their extension to quadratic fields, J. Number Theory 81 (2000) 210–233.
[2] J. Buddenhagen, C. Ford, M. May, Nice cubic polynomials, Pythagorean triples, and the law of
cosines, Math. Mag. 65 (1992) 244–249.
[3] C.K. Caldwell, Nice polynomials of degree 4, Math. Spectr. 23 (1990) 36–39.
[4] J.E. Cremona, Elliptic curve data, at website http://homepages.warwick.ac.uk/~masgaj/ftp/data/.
[5] L.E. Dickson, History of the Theory of Numbers, vol. 2, Chelsea Publishing Company, New York,
1992, reprint.
[6] J.C. Evard, Polynomials whose roots and critical points are integers, at the website, http://arxiv.
org/abs/math/0407256.
[7] S.K. Gupta, W. Szymanski, Cubic polynomials with rational roots and critical points, College Math.
J. 41 (2010) 365–369.