betaFeedback

WorksMathematics

Irreducibility of the Cuboid Polynomial Qp,q(t)Q_{p,q}(t) via the Rank-One Chabauty-Coleman Method

Abstract

Let Qp,q(t)∈Z[t]Q_{p,q}(t)\in\mathbb{Z}[t] be Sharipov's even monic degree-1010 second cuboid polynomial, with coprime positive integers p≠qp\ne q. We prove that Qp,q(t)Q_{p,q}(t) is irreducible over Q\mathbb{Q}, and hence over Z\mathbb{Z}. More generally, the normalized associated quintic Ps(x)P_s(x) and its composition Ps(t2)P_s(t^2) are irreducible over Q\mathbb{Q} for every s∈Q∖{0,±1}s\in\mathbb{Q}\setminus\{0,\pm1\}. The proof reduces linear and quadratic factors to rational points and horizontal conjugate pairs on one fixed genus-22 curve. A certified rank upper bound for its Jacobian, together with an exact divisor calculation, gives a rank-one annihilating differential at the prime 33. For the quadratic-factor case, the symmetric coordinates of a horizontal pair lie on an explicit double cover of an elliptic quartic. All rational affine points on this model are 33-adically integral and reduce to six smooth residue classes. The pulled-back differential is a unit in each class, proving that six displayed points form the complete list. Their coordinates exclude every admissible parameter. A negative real root of the irreducible quintic then supplies the additional degree argument required for irreducibility of the degree-1010 polynomial. The rank computation and all exact arithmetic certificates are specified explicitly for reproduction.

Introduction

The perfect cuboid problem asks for a rectangular box with positive integer edges whose three face diagonals and space diagonal are integers. In Sharipov’s polynomial formulation [1, 2], one encounters explicit even polynomials depending on integer parameters. The second cuboid polynomial Qp,q(t)Q_{p,q}(t) has degree 10 and is conjectured to be irreducible for coprime positive integers p≠qp \ne q.

Goal. We prove this irreducibility statement. The intermediate result is stronger than required by the original parametrization: the associated normalized quintic Ps(x)P_s(x) is irreducible for every s∈Q>0∖{1}s \in\mathbb{Q}_{>0} \setminus\{1\}, without assuming that ss is a rational square.

Method. A reducible quintic has an irreducible factor of degree at most two. An explicit transformation sends a root of PsP_s to a fixed genus-2 curve CC, with second coordinate (s−1)/(s+1)(s-1)/(s+1). A rational root gives a rational point of CC. An irreducible quadratic factor gives a conjugate pair on CC whose two points have the same second coordinate. Thus both cases are controlled by the same Jacobian.

We use the rank-one form of the Chabauty–Coleman method [3, 4]. The treatment of conjugate pairs is related to symmetric-power methods [5], but only a one-dimensional locus of horizontal pairs is needed here. We construct its model directly and prove the required local uniqueness rather than invoke a rational-point theorem for the entire symmetric square. A rank bound for the genus-2 Jacobian is obtained computationally, in the setting of explicit descent [6], using Magma [7]. All subsequent steps use a small exact divisor calculation, power series over Z3\mathbb{Z}_3, and enumeration over F3\mathbb{F}_3. In particular, the proof does not assume completeness of a bounded search for rational points.

The article is organized as follows. Section 2 states the normalization and the main theorem. Sections 3–4 establish the rank-one differential. Section 5 excludes linear factors. Sections 6–7 determine the horizontal-pair obstruction and exclude quadratic factors. Section 8 completes the degree-10 argument. The appendix records the computational inputs and reproducible verification code.

The second cuboid polynomial and normalization

Let p,q∈Z>0p,q \in\mathbb{Z}_{>0} be coprime and p≠qp \ne q. Sharipov’s second cuboid polynomial is

Qp,q(t)=t10+(2q2+p2)(3q2−2p2)t8+(q8+10p2q6+4p4q4−14p6q2+p8)t6−p2q2(q8−14p2q6+4p4q4+10p6q2+p8)t4−p6q6(q2+2p2)(−2q2+3p2)t2−p10q10.(1)\begin{aligned} Q_{p,q}(t)={}&t^{10}+(2q^2+p^2)(3q^2-2p^2)t^8\\ &+(q^8+10p^2q^6+4p^4q^4-14p^6q^2+p^8)t^6\\ &-p^2q^2(q^8-14p^2q^6+4p^4q^4+10p^6q^2+p^8)t^4\\ &-p^6q^6(q^2+2p^2)(-2q^2+3p^2)t^2-p^{10}q^{10}. \tag*{(1)} \end{aligned}

It is weighted-homogeneous of weight 20 for deg⁡p=deg⁡q=1\deg p=\deg q=1 and deg⁡t=2\deg t=2.

Definition 1 (Normalized second cuboid quintic). For s∈Qs \in\mathbb{Q}, define

Ps(x)=x5+(2+s)(3−2s)x4+(1+10s+4s2−14s3+s4)x3−s(1−14s+4s2+10s3+s4)x2−s3(1+2s)(−2+3s)x−s5.(2)P_s(x)=x^5+(2+s)(3-2s)x^4+(1+10s+4s^2-14s^3+s^4)x^3-s(1-14s+4s^2+10s^3+s^4)x^2-s^3(1+2s)(-2+3s)x-s^5. \tag*{(2)}

Lemma 1 (Normalization). With s=(p/q)2s=(p/q)^2, one has

Qp,q(t)=q20Ps(t2q4).(3)Q_{p,q}(t)=q^{20}P_s\left(\frac{t^2}{q^4}\right). \tag*{(3)}

Proof. Set r=p/qr=p/q and substitute p=rqp=rq, t=q2ut=q^2u into (1). Factoring out q20q^{20} and setting s=r2s=r^2 gives Ps(u2)P_s(u^2).

Theorem 1. For every s∈Q>0∖{1}s\in\mathbb{Q}_{>0}\setminus\{1\}, the quintic Ps(x)P_s(x) is irreducible over Q\mathbb{Q}. Consequently, for coprime positive integers p≠qp\ne q, the polynomial Qp,q(t)Q_{p,q}(t) is irreducible over Q\mathbb{Q} and over Z\mathbb{Z}.

The last implication requires a separate argument: irreducibility of Ps(x)P_s(x) alone does not automatically give irreducibility of Ps(t2)P_s(t^2). That argument will be supplied in Section 8. § 8

A fixed genus-two curve and a rank bound

Put

f(z)=−16z5+48z4−48z3+16z2+z,(4)f(z) = -16z^{5} + 48z^{4} - 48z^{3} + 16z^{2} + z, \tag*{(4)}
h(z)=1−f(z)=(1−z)(1−2z)2(1+4z−4z2).(5)h(z) = 1 - f(z) = (1-z)(1-2z)^{2}(1+4z-4z^{2}). \tag*{(5)}

Let CC denote the smooth projective curve associated with

C:y2=f(z),(6)C : y^{2} = f(z), \tag*{(6)}

and write J=Jac⁡(C)J = \operatorname{Jac}(C). The discriminant of ff is

disc⁡(f)=−216⋅43.(7)\operatorname{disc}(f) = -2^{16} \cdot43. \tag*{(7)}

Hence CC has genus 2 and good reduction at 3. It has a unique rational point ∞\infty at infinity. We also use the rational Weierstrass point T=(0,0)T = (0,0). A basis of holomorphic differentials is

ω0=dz2y,ω1=z dz2y.(8)\omega_{0} = \frac{dz}{2y}, \qquad\omega_{1} = \frac{z\,dz}{2y}. \tag*{(8)}

Both are integral differentials at the prime 3.

Lemma 2 (Computational rank bound). The Mordell–Weil rank of J(Q)J(\mathbb{Q}) is at most one.

Computer-assisted proof. Consider the genus-2 model

C0:w2=g(t):=t5+21t4+26t3+10t2+5t+1.(9)C_{0} : \quad w^{2} = g(t) := t^{5} + 21t^{4} + 26t^{3} + 10t^{2} + 5t + 1. \tag*{(9)}

The substitution

z=1t+1,y=w(t+1)3(10)z = \frac{1}{t+1}, \qquad y = \frac{w}{(t+1)^{3}} \tag*{(10)}

satisfies

(t+1)6f(1t+1)=g(t).(t+1)^{6} f\left(\frac{1}{t+1}\right) = g(t).

Its inverse is t=z−1−1t = z^{-1} - 1, w=y/z3w = y/z^{3} on the corresponding affine open sets. Thus it extends to an isomorphism of smooth projective curves over Q\mathbb{Q}, and hence of their Jacobians. Let J0=Jac⁡(C0)J_{0} = \operatorname{Jac}(C_{0}). A certified computation in Magma gives

rank⁡J0(Q)≤1.\operatorname{rank} J_{0}(\mathbb{Q}) \le1.

The class-group calculations use full Minkowski bounds, so this upper bound is unconditional and does not rely on the generalized Riemann hypothesis. The complete computation and its output are recorded in Appendix § 9. Since (10) induces an isomorphism J≃J0J \simeq J_{0} over Q\mathbb{Q}, their Mordell–Weil ranks are equal, proving the assertion.

Remark 1. Lemma 2 is the computational input to the proof. We do not need a complete Mordell–Weil basis or a saturation calculation. An explicit divisor below will have nonzero 3-adic logarithm, proving that the rank is exactly one and determining its annihilating differential to the required precision.

An exact rank-one differential at the prime three

For a holomorphic differential ω\omega, let log⁡ω\log_\omega denote its Abelian logarithm on J(Q3)J(\mathbb{Q}_3); see [8] for the integration framework. On a residue disk this logarithm is obtained by integrating the corresponding local power series. The logarithm kills torsion and is additive on divisor classes. Its compatibility with pullback will also be used for the horizontal-pair model.

Lemma 3 (A local uniqueness criterion). Let a residue disk have a parameter q∈3Z3q \in3\mathbb{Z}_3, centered at q=0q = 0. Suppose a differential on it is b(q) dqb(q)\,dq, where b(q)∈Z3[[q]]b(q) \in\mathbb{Z}_3[[q]] and b(0)∈Z3×b(0) \in\mathbb{Z}_3^\times. Then the primitive vanishing at q=0q = 0 has no other zero in this disk.

Proof. Write b(q)=∑j≥0bjqjb(q) = \sum_{j\geq0} b_jq^j. For q≠0q \ne0, its primitive divided by qq is

b0+∑j≥1bjqjj+1.b_0 + \sum_{j\geq1}\frac{b_jq^j}{j+1}.

For each j≥1j \geq1, one has v3(qj/(j+1))≥j−v3(j+1)≥1v_3(q^j/(j+1)) \geq j-v_3(j+1) \geq1. Thus this expression is a unit, proving the claim.

We represent divisor classes by Mumford coordinates (U,V)(U,V): here UU is monic, deg⁡V<deg⁡U≤2\deg V < \deg U \leq2, and U∣f−V2U \mid f - V^2. The corresponding class is the effective divisor supported on U(z)=0U(z) = 0, y=V(z)y = V(z), minus (deg⁡U)∞(\deg U)\infty. Such coordinates allow exact addition by Cantor’s algorithm [9].

Lemma 4 (Annihilating differential). There exists a differential annihilating J(Q)J(\mathbb{Q}) of the form

ω=aω0+ω1,a∈Z3,a≡3(mod9).(11)\omega= a\omega_0 + \omega_1,\qquad a \in\mathbb{Z}_3,\qquad a \equiv3\pmod9. \tag*{(11)}

Proof. Let

A=[(1/2,1)−∞]∈J(Q).A = \left[(1/2,1)-\infty\right] \in J(\mathbb{Q}).

Exact divisor addition gives the following small certificate:

11A=(U,V),U(z)=z2−69196z−9392,V(z)=−40852058z−2331372.(12)\begin{aligned} 11A &= (U,V),\\ U(z) &= z^2-\frac{69}{196}z-\frac{9}{392},\\ V(z) &= -\frac{4085}{2058}z-\frac{233}{1372}. \tag*{(12)} \end{aligned}

Appendix § 9 specifies checks for both the divisor-class equality and the divisibility U∣f−V2U \mid f - V^2. The latter condition alone would not certify that the class is 11A11A.

Let (zi,yi)(z_i,y_i), i=1,2i = 1,2, be the two geometric points represented by (12). Since yi=V(zi)y_i = V(z_i), their trace and norm are obtained directly from UU:

τ:=y1+y2=−139623134456,ν:=y1y2=7671134456.(13)\tau:= y_1+y_2 = -\frac{139623}{134456},\qquad\nu:= y_1y_2 = \frac{7671}{134456}. \tag*{(13)}

Write τ=3r\tau= 3r and ν=3n\nu= 3n. Then

r,n∈Z3×,r≡n≡2(mod3).(14)r,n \in\mathbb{Z}_3^\times,\qquad r \equiv n \equiv2\pmod3. \tag*{(14)}

The nonleading coefficients of UU have valuations 1 and 2. Its Newton polygon therefore gives v3(zi)=1v_3(z_i)=1. From f(z)=z+O(z2)f(z)=z+O(z^2) it follows that v3(yi)=1/2v_3(y_i)=1/2. Both geometric points lie in the disk of TT over a finite extension of Q3\mathbb{Q}_3.

Near TT, the coordinate yy is a uniformizer and the equation y2=f(z)y^2=f(z) has inverse series

z=y2−16y4+560y6+⋯∈Z3[[y2]].(15)z=y^2-16y^4+560y^6+\cdots\in\mathbb{Z}_3[[y^2]]. \tag*{(15)}

The integrality follows by formal inversion, because f′(0)=1f'(0)=1. Since div⁡(z)=2T−2∞\operatorname{div}(z)=2T-2\infty, the class [T−∞][T-\infty] is 2-torsion and [2T−2∞]=0[2T-2\infty]=0. Consequently, with Sm=y1m+y2mS_m=y_1^m+y_2^m, the logarithms of 11A11A are sums of integrals from TT:

ℓ0:=log⁡ω0(11A)=τ−323S3+⋯ ,(16)\ell_0:=\log_{\omega_0}(11A)=\tau-\frac{32}{3}S_3+\cdots, \tag*{(16)}
ℓ1:=log⁡ω1(11A)=13S3−485S5+⋯ .(17)\ell_1:=\log_{\omega_1}(11A)=\frac{1}{3}S_3-\frac{48}{5}S_5+\cdots. \tag*{(17)}

These formulas result by differentiating (15), substituting in (8), and integrating term by term. Both local integrals are computed in a finite extension K/Q3K/\mathbb{Q}_3 containing the two support points, with v3(3)=1v_3(3)=1. Since the unintegrated coefficients are integral and v3(yi)=1/2v_3(y_i)=1/2, a term of degree 2j−12j-1 in either primitive has valuation at least j−12−v3(2j−1)j-\frac{1}{2}-v_3(2j-1), which tends to infinity. Thus the individual series converge in KK and may be added term by term to obtain the expressions in S2j−1S_{2j-1}.

We make the precision bounds explicit. The power sums satisfy S0=2S_0=2, S1=τS_1=\tau, and

Sm=τSm−1−νSm−2.S_m=\tau S_{m-1}-\nu S_{m-2}.

For m=2j−1m=2j-1, each monomial in the expression for SmS_m in terms of τ,ν\tau,\nu contains at least jj factors, so

v3(S2j−1)≥j.(18)v_3(S_{2j-1})\ge j. \tag*{(18)}

Moreover,

S3=τ3−3ντ=27r(r2−n).(19)S_3=\tau^3-3\nu\tau=27r(r^2-n). \tag*{(19)}

All unintegrated power-series coefficients are integral; integration introduces only the denominator 2j−12j-1 in the term involving S2j−1S_{2j-1}. In (16), the S3S_3 term therefore has valuation at least 2, and every subsequent term also has valuation at least 2. In (17), all terms after S3S_3 have valuation at least 3, since

j−v3(2j−1)≥3(j≥3).j-v_3(2j-1)\ge3 \qquad(j\ge3).

The valuations tend to infinity, so these estimates also justify passage to the infinite sums. It follows that

ℓ0≡3r(mod9),ℓ1≡9r(r2−n)(mod27).(20)\ell_0\equiv3r\pmod{9},\qquad\ell_1\equiv9r(r^2-n)\pmod{27}. \tag*{(20)}

In particular, ℓ0≠0\ell_0\ne0, so AA has infinite order. By Lemma 2, the group J(Q)J(\mathbb{Q}) has rank exactly one. Its logarithmic image spans the same Q3\mathbb{Q}_3-line as log⁡(11A)\log(11A).

Set a=−ℓ1/ℓ0a=-\ell_1/\ell_0. This gives an annihilating differential for every rational divisor class. The congruences above imply a∈3Z3a\in3\mathbb{Z}_3 and

a3≡−(r2−n)≡1(mod3).\frac{a}{3}\equiv-(r^2-n)\equiv1\pmod{3}.

Thus a≡3(mod9)a\equiv3\pmod{9}, as required. The argument only uses a nonzero vector in the rank-one logarithmic image; no assertion that AA generates the full Mordell–Weil group is needed.

Rational points and exclusion of linear factors

Proposition 1 (Rational points on CC). The full rational-point set is

C(Q)={∞,(0,0),(1,±1),(1/2,±1)}.(21)C(\mathbb{Q}) = \{\infty,(0,0),(1,\pm1),(1/2,\pm1)\}. \tag*{(21)}

Proof. Enumeration over F3\mathbb{F}_3 gives

C(F3)={∞,(0,0),(1,±1),(2,±1)}.(22)C(\mathbb{F}_3) = \{\infty,(0,0),(1,\pm1),(2,\pm1)\}. \tag*{(22)}

Every rational point reduces to one of these six points by properness of the smooth model. At each of the four finite points with nonzero first coordinate, the reduction of ω\omega is the nonzero differential z dz/(2y)z\,dz/(2y). It is also nonzero at infinity. Indeed, the local parameter z2/yz^2/y at infinity shows that z dz/(2y)z\,dz/(2y) has unit leading coefficient, whereas dz/(2y)dz/(2y) vanishes there. Each of these five disks contains exactly one of the displayed rational points. The Abelian integral of ω\omega between rational points vanishes, so Lemma 3 gives uniqueness in each disk.

It remains to examine the disk of T=(0,0)T=(0,0), where the reduced differential vanishes. Use y∈3Z3y\in3\mathbb{Z}_3 as parameter and put

FT(y)=∫T(z(y),y)ω.F_T(y)=\int_T^{(z(y),y)}\omega.

From (15) one obtains, for y≠0y\ne0,

FT(y)y=a+1−32a3y2+R(y),v3(R(y))≥4.(23)\frac{F_T(y)}{y}=a+\frac{1-32a}{3}y^2+R(y),\qquad v_3(R(y))\ge4. \tag*{(23)}

For the remainder bound, the terms remaining after division by yy have degree 2k2k, k≥2k\ge2, and their coefficients belong to (2k+1)−1Z3(2k+1)^{-1}\mathbb{Z}_3. For y∈3Z3y\in3\mathbb{Z}_3 use 2k−v3(2k+1)≥42k-v_3(2k+1)\ge4 to obtain the stated valuation bound.

If v3(y)=1v_3(y)=1, write y=3cy=3c with c∈Z3×c\in\mathbb{Z}_3^\times. Using a/3≡1(mod3)a/3\equiv1\pmod3 gives

FT(y)3y≡1+c2≡2(mod3).\frac{F_T(y)}{3y}\equiv1+c^2\equiv2\pmod3.

If v3(y)≥2v_3(y)\ge2, the term aa on the right of (23) has valuation one, whereas every other term has larger valuation. Hence FT(y)≠0F_T(y)\ne0 whenever y≠0y\ne0. Since the integral must vanish at any rational point in this disk, TT is its only rational point. This completes the list.

We now connect this fixed curve to the parameter-dependent quintic. For s≠0s\ne0, define the reciprocal polynomial

Rs(x)=−x5s5Ps(s2x)∈Q[x].(24)R_s(x)=-\frac{x^5}{s^5}P_s\left(\frac{s^2}{x}\right)\in\mathbb{Q}[x]. \tag*{(24)}

Its leading coefficient is one. Put

Δs(x)=(x+1)(x+s2),Zs(x)=x(s−1)2Δs(x),vs=s−1s+1.(25)\Delta_s(x)=(x+1)(x+s^2),\qquad Z_s(x)=\frac{x(s-1)^2}{\Delta_s(x)},\qquad v_s=\frac{s-1}{s+1}. \tag*{(25)}

A direct polynomial identity is

(s+1)2Δs(x)5(f(Zs(x))−vs2)=−(s−1)2Ps(x)Rs(x).(26)(s+1)^2\Delta_s(x)^5\left(f(Z_s(x))-v_s^2\right)=-(s-1)^2P_s(x)R_s(x). \tag*{(26)}

Both sides become polynomials in s,xs,x after clearing the displayed substitution; Appendix A includes its exact verification.

The transformation can also be motivated by inversion invariants. With ρ=x/s\rho=x/s, set U=s+1/sU=s+1/s and V=ρ+1/ρV=\rho+1/\rho. The quotient parameter

t=V+2U−2=(x+s)2x(s−1)2t=\frac{V+2}{U-2}=\frac{(x+s)^2}{x(s-1)^2}

gives z=1/(t+1)=Zs(x)z=1/(t+1)=Z_s(x), in agreement with (10). For the proof itself, the identity (26) is sufficient.

Lemma 5 (Roots give points of CC). For s∈Q>0∖{1}s\in\mathbb{Q}_{>0}\setminus\{1\}, every root xx of PsP_s gives a point (Zs(x),vs)(Z_s(x),v_s) on CC, defined over Q(x)\mathbb{Q}(x). Its first coordinate is finite and nonzero.

Proof. The relevant exceptional values are

Ps(0)=−s5,Ps(−1)=4(s−1)4(s+1),Ps(−s2)=−4s5(s−1)4(s+1).(27)\begin{aligned} P_s(0)&=-s^5,\\ P_s(-1)&=4(s-1)^4(s+1),\\ P_s(-s^2)&=-4s^5(s-1)^4(s+1). \tag*{(27)} \end{aligned}

They are nonzero under the stated assumptions. Thus Zs(x)Z_s(x) is defined and nonzero at every root. The identity (26) now gives f(Zs(x))=vs2f(Z_s(x))=v_s^2.

Corollary 1 (No linear factor). For s∈Q>0∖{1}s\in\mathbb{Q}_{>0}\setminus\{1\}, the polynomial PsP_s has no rational root.

Proof. A rational root would give a rational point of CC whose second coordinate is vsv_s. But −1<vs<1-1<v_s<1 and vs≠0v_s\ne0, whereas Proposition 1 permits only second coordinates 0,1,−10,1,-1 at finite rational points.

Quadratic factors and horizontal conjugate pairs

We first check that distinct roots of PsP_s cannot coalesce under the map ZsZ_s. This matters because dividing a difference of curve equations by the difference of their first coordinates will be essential.

Lemma 6 (No collision of root images). For s∈Q>0∖{1}s\in\mathbb{Q}_{>0}\setminus\{1\}, two distinct roots x1,x2x_1,x_2 of PsP_s have distinct images under ZsZ_s.

Proof. Coefficient comparison in (2) and (24) gives

Ps(x)−Rs(x)=8(1−s2)x(x+s)3,Ps(−s)=−2s3(s−1)4.(28)P_s(x)-R_s(x)=8(1-s^2)x(x+s)^3,\qquad P_s(-s)=-2s^3(s-1)^4. \tag*{(28)}

Thus PsP_s cannot have both xx and s2/xs^2/x as roots. Indeed, such an xx would be a common root of Ps,RsP_s,R_s, forcing x=0x=0 or x=−sx=-s, and neither is a root. On the other hand,

Zs(x1)−Zs(x2)=(s−1)2(x1−x2)(s2−x1x2)Δs(x1)Δs(x2).(29)Z_s(x_1)-Z_s(x_2)=\frac{(s-1)^2(x_1-x_2)(s^2-x_1x_2)}{\Delta_s(x_1)\Delta_s(x_2)}. \tag*{(29)}

All denominators are nonzero by (27). Equality of the images would therefore force x1x2=s2x_1x_2=s^2, contrary to the preceding observation.

Suppose PsP_s has a quadratic divisor over Q\mathbb{Q}. By Corollary 1 this divisor is irreducible, so its roots x1,x2x_1,x_2 are distinct quadratic conjugates. By Lemmas 5 and 6, their images

(z1,vs), (z2,vs)∈C(Q‾)(z_1,v_s),\ (z_2,v_s) \in C(\overline{\mathbb{Q}})

are distinct conjugate points. Their common second coordinate makes this a horizontal conjugate pair. Write

e=z1+z2,m=z1z2,u=e−1,Y=4m−6u2.(30)e=z_1+z_2,\qquad m=z_1z_2,\qquad u=e-1,\qquad Y=4m-6u^2. \tag*{(30)}

All four quantities are rational. Since h(z1)=h(z2)h(z_1)=h(z_2), division by z1−z2z_1-z_2 gives

16m2−48(e−1)2m+16e(e−1)3−1=0.(31)16m^2-48(e-1)^2m+16e(e-1)^3-1=0. \tag*{(31)}

After completing the square in mm, this becomes the quartic

E:Y2=20u4−16u3+1.(32)E:\qquad Y^2=20u^4-16u^3+1. \tag*{(32)}

The smooth projective completion of this nonsingular quartic, with the rational point (0,1)(0,1) chosen as origin, is an elliptic curve.

To recover the common value of hh, reduce it modulo z2−ez+mz^2-ez+m. Using (31), the remainder is constant and equals

W=1−16mu3+16m2(2u−1)=2u(10Yu2−6Yu+44u4−44u3+8u2+1),(33)\begin{aligned} W&=1-16mu^3+16m^2(2u-1)\\ &=2u(10Yu^2-6Yu+44u^4-44u^3+8u^2+1), \tag*{(33)} \end{aligned}

where m=(Y+6u2)/4m=(Y+6u^2)/4 in the second line. Since f=1−hf=1-h, the horizontal pair gives a rational point of the affine model

X:Y2=20u4−16u3+1,v2=1−W(u,Y),(34)X:\qquad Y^2=20u^4-16u^3+1,\qquad v^2=1-W(u,Y), \tag*{(34)}

with v=vsv=v_s. Thus the quadratic-factor problem has become a rational-point problem on a fixed curve.

Lemma 7 (Divisor map and pulled-back differentials). For a point of XX, put

e=u+1,m=Y+6u24,B(z)=z2−ez+m.e=u+1,\qquad m=\frac{Y+6u^2}{4},\qquad B(z)=z^2-ez+m.

The two roots of BB define an effective degree-two divisor

D=(z1,v)+(z2,v)D=(z_1,v)+(z_2,v)

on CC, with multiplicities when the roots coincide. The class Φ=[D−2∞]\Phi=[D-2\infty] is rational at rational points of XX. On the locus v≠0v\ne0, the pullbacks of (8) are

η0=du2v,η1=(1−2u) du−14dY2v.(35)\eta_0=\frac{du}{2v},\qquad\eta_1=\frac{(1-2u)\,du-\frac{1}{4}dY}{2v}. \tag*{(35)}

Proof. The reduction calculation above gives B∣h−WB \mid h-W. By (34), it follows that B∣f−v2B \mid f-v^2, so the indicated horizontal divisor lies on CC. Its definition is Galois invariant when uu, YY, vv are rational. The map to the Jacobian is given by its degree-zero divisor class.

For distinct roots of BB, the pullback of a differential to this class is the sum of its values on the two points. Since z1+z2=ez_1+z_2=e and z1z2=mz_1z_2=m,

dz1+dz2=de=du,z1dz1+z2dz2=e de−dm.dz_1+dz_2=de=du,\qquad z_1dz_1+z_2dz_2=e\,de-dm.

Now dm=14dY+3u dudm=\frac{1}{4}dY+3u\,du, which gives (35). The formulas are symmetric in the roots and extend across coincident roots wherever v≠0v\ne0.

The differential

η=aη0+η1(36)\eta=a\eta_0+\eta_1 \tag*{(36)}

is the pullback of the annihilator in Lemma 4. For rational points RR, R0R_0 of XX in a common smooth residue disk on which these expressions are integral, its local integral satisfies

∫R0Rη=log⁡ω(Φ(R)−Φ(R0))=0.(37)\int_{R_0}^{R}\eta=\log_{\omega}\bigl(\Phi(R)-\Phi(R_0)\bigr)=0. \tag*{(37)}

In the disks used below, YY and vv are units and all denominators are units at 3. The divisor construction and the displayed differentials are therefore defined integrally on each disk.

Six residue disks determine the obstruction points

Proposition 2 (Complete affine rational-point list). The rational affine points of (34) are exactly

(u,Y,v)=(0,1,±1),(0,−1,±1),(1/2,1/2,±1).(38)(u,Y,v)=(0,1,\pm1),\qquad(0,-1,\pm1),\qquad(1/2,1/2,\pm1). \tag*{(38)}

Proof. First, every rational affine point is integral at 3. If v3(u)<0v_3(u)<0, divide (32) by u4u^4:

(Yu2)2=20−16u+1u4.\left(\frac{Y}{u^2}\right)^2=20-\frac{16}{u}+\frac{1}{u^4}.

The right side is a unit with reduction 2 in F3\mathbb{F}_3, which cannot be the reduction of a square in Q3\mathbb{Q}_3. Hence u∈Z3u\in\mathbb{Z}_3. The two equations of XX then give Y,v∈Z3Y,v\in\mathbb{Z}_3. Consequently every rational affine point has a reduction in this affine model.

Enumerating the equations over F3\mathbb{F}_3 yields precisely

(u,Y,v)=(0,1,±1),(0,2,±1),(2,2,±1).(39)(u,Y,v)=(0,1,\pm1),\qquad(0,2,\pm1),\qquad(2,2,\pm1). \tag*{(39)}

At all six points both YY and vv are nonzero. The derivatives of the two equations with respect to Y,vY,v form an invertible triangular matrix, with diagonal entries 2Y2Y and 2v2v. Thus all six points are smooth, and u−u0u-u_0 is a local parameter in the disk of any chosen lift (u0,Y0,v0)(u_0,Y_0,v_0). Each disk contains one of the six rational points in (38); direct substitution gives W=0W=0 at those points.

By Lemma 4, a≡0(mod3)a \equiv0 \pmod{3}, so η\eta reduces to η1\eta_1. Differentiating (32) gives

dY=80u3−48u22Y du.dY = \frac{80u^3-48u^2}{2Y}\,du.

Therefore

η1=1−2u−(80u3−48u2)/(8Y)2v du.(40)\eta_1 = \frac{1-2u-(80u^3-48u^2)/(8Y)}{2v}\,du. \tag*{(40)}

The numerator reduces to 1 at either point with u=0u=0, and to 2 at (u,Y)=(2,2)(u,Y)=(2,2). Since vv is a unit, the coefficient of dudu is a unit in every disk in (39).

Fix the known rational point R0R_0 in one of these disks. Any other rational point RR in the disk satisfies (37). Lemma 3, applied with parameter u−u(R0)u-u(R_0), forces R=R0R=R_0. Every rational affine point belongs to one of the six disks, so the displayed list is complete.

Corollary 2 (No quadratic factor). For s∈Q>0∖{1}s \in\mathbb{Q}_{>0} \setminus\{1\}, the polynomial PsP_s has no quadratic divisor over Q\mathbb{Q}.

Proof. The construction of Section 6 sends any such divisor to a rational affine point of XX with v=(s−1)/(s+1)v=(s-1)/(s+1). For positive ss this satisfies −1<v<1-1<v<1. Proposition 2 gives v=±1v=\pm1 at every rational affine point of XX, a contradiction.

Remark 2. The integrality argument is essential: it ensures that the six affine residue disks cover every rational point required in the factorization argument. No bound on the numerator or denominator of uu is used. The uniqueness in each disk then proves completeness, rather than merely confirming a list found by search.

Completion of the irreducibility proof

Proof of Theorem 1. A reducible polynomial of degree five over a field has an irreducible factor of degree at most two. Indeed, if every irreducible factor had degree at least three, two nonconstant factors would already have total degree at least six. Corollaries 1 and 2 therefore prove irreducibility of Ps(x)P_s(x) over Q\mathbb{Q}.

We next establish irreducibility after composition with a square. By (27),

Ps(−1)=4(s−1)4(s+1)>0,Ps(0)=−s5<0.P_s(-1)=4(s-1)^4(s+1)>0,\qquad P_s(0)=-s^5<0.

Thus PsP_s has a real root α∈(−1,0)\alpha\in(-1,0). Its irreducibility gives [Q(α):Q]=5[\mathbb{Q}(\alpha):\mathbb{Q}]=5. View Q(α)\mathbb{Q}(\alpha) as the subfield of R\mathbb{R} defined by this negative real root. A square in this real field is nonnegative, so α\alpha is not a square in Q(α)\mathbb{Q}(\alpha). The polynomial u2−αu^2-\alpha is consequently irreducible over Q(α)\mathbb{Q}(\alpha).

If β2=α\beta^2=\alpha, then Q(α)⊂Q(β)\mathbb{Q}(\alpha)\subset\mathbb{Q}(\beta) and

[Q(β):Q]=[Q(β):Q(α)][Q(α):Q]=2⋅5=10.[\mathbb{Q}(\beta):\mathbb{Q}]=[\mathbb{Q}(\beta):\mathbb{Q}(\alpha)][\mathbb{Q}(\alpha):\mathbb{Q}]=2\cdot5=10.

Since β\beta is a root of the monic degree-10 polynomial Ps(u2)P_s(u^2), that polynomial is irreducible over Q\mathbb{Q}. For s=(p/q)2s=(p/q)^2, the invertible rational change of variable u=t/q2u=t/q^2 and the nonzero scalar factor in (3) preserve irreducibility. Hence Qp,q(t)Q_{p,q}(t) is irreducible over Q\mathbb{Q}. Finally, it is monic with integer coefficients, so Gauss’s lemma gives irreducibility over Z\mathbb{Z}.

Corollary 3 (Classification of rational parameters). For s∈Qs \in\mathbb{Q}, both Ps(x)P_s(x) and Ps(t2)P_s(t^2) are irreducible over Q\mathbb{Q} if and only if s∉{0,±1}s \notin\{0,\pm1\}.

Proof. For s∉{0,±1}s \notin\{0,\pm1\}, the proofs of Lemmas 5 and 6 apply unchanged: all exceptional values and denominators used there remain nonzero. Moreover, vs=(s−1)/(s+1)∉{0,±1}v_s=(s-1)/(s+1) \notin\{0,\pm1\}. Propositions 1 and 2 therefore exclude linear and quadratic factors of PsP_s, proving its irreducibility.

For s>0s>0, irreducibility of Ps(t2)P_s(t^2) follows from the degree argument above. For s<0s<0, we have Ps(0)>0P_s(0)>0 and Ps(x)→−∞P_s(x)\to-\infty as x→−∞x\to-\infty, so a negative real root exists and the same argument applies. Finally,

P0(x)=x3(x2+6x+1),P_0(x)=x^3(x^2+6x+1),
P1(x)=(x−1)(x+1)4,P_1(x)=(x-1)(x+1)^4,
P−1(x)=(x+1)5,P_{-1}(x)=(x+1)^5,

which proves reducibility in all exceptional cases, also after substituting x=t2x=t^2.

Magma scripts and transcripts

The only rank computation used in the proof is the upper bound in Lemma 2. Except for the auxiliary genus-two rational-point verification, the remaining computational checks use exact rational polynomial arithmetic or finite enumeration over F3\mathbb{F}_3. The code below is arranged to keep these two roles separate. Its rank target is 1; the remaining assertions check the isomorphism of curve models, the root-map and reciprocal identities, the divisor certificate, its trace and norm, and the residue-class calculations. The analytic uniqueness arguments are proved in the main text and do not depend on a rational-point search.

Script: Complete exact verification

The script was executed with Magma V2.29-10. The two class-group bound maps are both set to the Minkowski bound, so the rank calculation is unconditional. The additional Chabauty call independently reproduces the genus-two rational-point list; the proofs in the main text require only the rank upper bound and the exact local certificate. The identities and finite-field checks are assertions, so an unsuccessful check must be resolved before treating the run as a certificate.

In the final part of the script, the variables e,f,w denote the symmetric coordinates (e,m,v)(e,m,v). They are related to the coordinates of the main text by

u=e−1,Y=4m−6(e−1)2,v=w;u=e-1,\qquad Y=4m-6(e-1)^2,\qquad v=w;

the auxiliary variable V in that part equals Y/2Y/2. This change of coordinates is invertible over Z3\mathbb{Z}_3. Consequently, the printed rational lifts and residue classes give exactly the lists in (38) and (39), respectively, in symmetric coordinates. The nonvanishing checks use these same coordinates and verify the unit condition for the pulled-back differential.

Code.

Q := Rationals();
Rt<t> := PolynomialRing(Q);
F0 := t^5+21*t^4+26*t^3+10*t^2+5*t+1;
C0 := HyperellipticCurve(F0);
J0 := Jacobian(C0);
proof_bound := map<PowerStructure(RngOrd) -> Integers() |
    O :-> MinkowskiBound(O)>;
SetClassGroupBoundMaps(proof_bound, proof_bound);
upper := RankBound(J0);
print "Unconditional rank upper bound:", upper;
assert upper eq 1;
// Independent exact verification of the genus-two rational-point list.
B0 := J0![C0![0,1,1], C0![1,0,0]];
assert Order(B0) eq 0;
points0 := Chabauty(B0);
expected0 := {C0![1,0,0], C0![-1,0,1], C0![0,1,1],
    C0![0,-1,1], C0![1,8,1], C0![1,-8,1]};
assert #points0 eq 6;
assert &and[p in points0 : p in expected0];
print "Complete genus-two rational-point list:", points0;
Rz<z> := PolynomialRing(Q);
F := -16*z^5+48*z^4-48*z^3+16*z^2+z;
assert Discriminant(F) eq -2^16*43;
Kz := FieldOfFractions(Rz);
assert Kz!F eq Evaluate(F0, 1/(Kz!z)-1)*(Kz!z)^6;
C := HyperellipticCurve(F);
J := Jacobian(C);
// Exact 11A certificate; A denotes [(1/2,1)-infinity].
A := J![C![1/2,1,1], C![1,0,0]];
M := 11*A;
u := z^2-(69/196)*z-9/392;
v := -(4085/2058)*z-233/1372;
assert M[1] eq u and M[2] eq v and M[3] eq 2;
assert (F-v^2) mod u eq 0;
trace_y := -Coefficient(u,1)*Coefficient(v,1)+2*Coefficient(v,0);
norm_y := Coefficient(v,0)^2
    -Coefficient(u,1)*Coefficient(v,0)*Coefficient(v,1)
    +Coefficient(u,0)*Coefficient(v,1)^2;
assert trace_y eq -139623/134456;
assert norm_y eq 7671/134456;
assert Valuation(trace_y,3) eq 1 and Valuation(norm_y,3) eq 1;
trace_y3 := trace_y^3-3*norm_y*trace_y;
assert Valuation(trace_y3,3) eq 3;
function Residue(q,n)
    R := Integers(n);
    return (R!Numerator(q))/(R!Denominator(q));
end function;
assert [Residue(Coefficient(u,i),81) : i in [2,1,0]] eq [1,48,63];
assert [Residue(3*Coefficient(v,i),81) : i in [1,0]] eq [14,75];
assert Residue(trace_y/3,3) eq 2;
assert Residue(norm_y/3,3) eq 2;
assert Residue(trace_y3/27,3) eq 1;
print "Exact 11A Mumford certificate:", M;
print "Trace and norm:", trace_y, norm_y;
print "Scaled residues S/3, N/3, Tr(y^3)/27 modulo 3:",
    Residue(trace_y/3,3), Residue(norm_y/3,3), Residue(trace_y3/27,3);
// First terms of the integral formal inverse and local primitives.
RY<Y> := PolynomialRing(Q);
inverse_z := Y^2-16*Y^4+560*Y^6;
assert (Evaluate(F,inverse_z)-Y^2) mod Y^8 eq 0;
omega0 := RY!(Derivative(inverse_z)/(2*Y));
omega1 := (inverse_z*omega0) mod Y^6;
assert Derivative(Y-(32/3)*Y^3+336*Y^5) eq omega0;
assert Derivative((1/3)*Y^3-(48/5)*Y^5) eq omega1;
assert -Residue(trace_y3/27,3)/Residue(trace_y/3,3) eq 1;
// Symbolic polynomial/root-map identities over Q(s).
Ks<s> := FunctionField(Q);
Rx<x> := PolynomialRing(Ks);
P := x^5+(2+s)*(3-2*s)*x^4
    +(1+10*s+4*s^2-14*s^3+s^4)*x^3
    -s*(1-14*s+4*s^2+10*s^3+s^4)*x^2
    -s^3*(1+2*s)*(-2+3*s)*x-s^5;
Kx := FieldOfFractions(Rx);
X := Kx!x;
den := (X+1)*(X+s^2);
Z := X*(s-1)^2/den;
W0 := (s-1)/(s+1);
Pstar := Rx!(-X^5*Evaluate(P,s^2/X)/s^5);
assert (Evaluate(F,Z)-W0^2)*den^5*(s+1)^2
    eq -(s-1)^2*(Kx!P)*(Kx!Pstar);
assert Evaluate(P,-1) eq 4*(s-1)^4*(s+1);
assert Evaluate(P,-s^2) eq -4*s^5*(s-1)^4*(s+1);
assert Evaluate(P,0) eq -s^5;
assert P-Pstar eq 8*(1-s^2)*x*(x+s)^3;
assert Evaluate(P,-s) eq -2*s^3*(s-1)^4;
// Distinct roots have the same Z-image only if their product is s^2.
// The reciprocal identities above exclude that case.
Rpair<xx,yy> := PolynomialRing(Ks,2);
assert xx*(yy+1)*(yy+s^2)-yy*(xx+1)*(xx+s^2)
    eq (xx-yy)*(s^2-xx*yy);
// The equal-y divisor curve and its elliptic quotient.
B<e,f,w> := PolynomialRing(Q,3);
H := 16*f^2-48*(e-1)^2*f+16*e*(e-1)^3-1;
G := w^2-16*f*((e-1)^3-(2*e-3)*f);
KB := FieldOfFractions(B);
Rzz<zz> := PolynomialRing(KB);
remainder := Evaluate(F,zz) mod (zz^2-e*zz+f);
assert Coefficient(remainder,1) eq -KB!H;
assert Coefficient(remainder,0)-(KB!w)^2 eq -KB!G;
V := 2*f-3*(e-1)^2;
assert V^2-((e-1)^3*(5*e-9)+1/4) eq H/4;
// Exact known rational lifts, and all six affine F_3 residue classes.
lifts := [<Q!3/2,Q!1/2,Q!1>, <Q!3/2,Q!1/2,Q!-1>,
    <Q!1,Q!1/4,Q!1>, <Q!1,Q!1/4,Q!-1>,
    <Q!1,Q!-1/4,Q!1>, <Q!1,Q!-1/4,Q!-1>];
for p in lifts do
    assert Evaluate(H,[p[1],p[2],p[3]]) eq 0;
    assert Evaluate(G,[p[1],p[2],p[3]]) eq 0;
end for;
assert F-1 eq -(z-1)*(2*z-1)^2*(4*z^2-4*z-1);
F3 := GF(3);
B3<e3,f3,w3> := PolynomialRing(F3,3);
H3 := 16*f3^2-48*(e3-1)^2*f3+16*e3*(e3-1)^3-1;
G3 := w3^2-16*f3*((e3-1)^3-(2*e3-3)*f3);
points3 := {<ee,ff,ww> : ee,ff,ww in F3 |
    Evaluate(H3,[ee,ff,ww]) eq 0 and Evaluate(G3,[ee,ff,ww]) eq 0};
expected3 := {<F3!0,F3!2,F3!1>, <F3!0,F3!2,F3!2>,
              <F3!1,F3!1,F3!1>, <F3!1,F3!1,F3!2>,
              <F3!1,F3!2,F3!1>, <F3!1,F3!2,F3!2>};
assert points3 eq expected3;
assert {<F3!p[1],F3!p[2],F3!p[3]> : p in lifts} eq points3;
He := Derivative(H3,1);
Hf := Derivative(H3,2);
for p in points3 do
    he := Evaluate(He,[p[1],p[2],p[3]]);
    hf := Evaluate(Hf,[p[1],p[2],p[3]]);
    assert hf ne 0 and p[3] ne 0;
    assert p[1]+he/hf ne 0;
end for;
print "Six residue classes on the equal-y divisor curve:", points3;
print "Six rational lifts:", lifts;
print "Nonvanishing differential checks: passed in all six disks.";
print "ALL STANDALONE CERTIFICATE ASSERTIONS PASSED.";

Transcript

Unconditional rank upper bound: 1
Complete genus-two rational-point list: { (1 : -8 : 1), (0 : -1 : 1), (1 : 8 : 1), (-1 : 0 : 1), (0 : 1 : 1), (1 : 0 : 0) }
Exact 11A Mumford certificate: (x^2 - 69/196*x - 9/392, -4085/2058*x - 233/1372, 2)
Trace and norm: -139623/134456 7671/134456
Scaled residues S/3, N/3, Tr(y^3)/27 modulo 3: 2 2 1
Six residue classes on the equal-y divisor curve: { <0, 2, 1>, <1, 2, 2>, <0, 2, 2>, <1, 1, 1>, <1, 2, 1>, <1, 1, 2> }
Six rational lifts: [ <3/2, 1/2, 1>, <3/2, 1/2, -1>, <1, 1/4, 1>, <1, 1/4, -1>, <1, -1/4, 1>, <1, -1/4, -1> ]
Nonvanishing differential checks: passed in all six disks.
ALL STANDALONE CERTIFICATE ASSERTIONS PASSED.

The auxiliary primitive calculations in the script use the same normalization ω0=dz/(2y)\omega_0 = dz/(2y) and ω1=z dz/(2y)\omega_1 = z\,dz/(2y) as the main text. The power-series assertions check their first coefficients; the all-orders precision estimates are proved in Lemma 4 and Proposition 1.

AI Assistance

OpenAI GPT-6.1 Sol assisted with the development of the argument, symbolic calculations, verification code, and the preparation of the exposition. The author is responsible for the mathematical content and the final presentation.

References

  1. [1]R. A. Sharipov, Perfect cuboids and irreducible polynomials, Ufa Math. J. 4 (2012), no. 1, 153–160.arxiv.org/abs/1108.5348
  2. [2]R. A. Sharipov, A note on the second cuboid conjecture. Part I, arXiv:1201.1229 [math.NT] (2012).arxiv.org/abs/1201.1229
  3. [3]C. Chabauty, Sur les points rationnels des courbes algébriques de genre supérieur à l’unité, C. R. Acad. Sci. Paris 212 (1941), 882–885.
  4. [4]R. F. Coleman, Effective Chabauty, Duke Math. J. 52 (1985), no. 3, 765–770.DOI
  5. [5]S. Siksek, Chabauty for symmetric powers of curves, Algebra Number Theory 3 (2009), no. 2, 209–236.DOI
  6. [6]M. Stoll, Implementing 2-descent for Jacobians of hyperelliptic curves, Acta Arith. 98 (2001), no. 3, 245–277.DOI
  7. [7]W. Bosma, J. Cannon, C. Playoust, The Magma Algebra System I: The User Language, J. Symbolic Computation 24 (1997), 235–265; see also https://magma.maths.usyd.edu.au/magma/.DOI
  8. [8]R. F. Coleman, Torsion points on curves and p-adic Abelian integrals, Ann. of Math. (2) 121 (1985), no. 1, 111–168.DOI
  9. [9]D. G. Cantor, Computing in the Jacobian of a hyperelliptic curve, Math. Comp. 48 (1987), no. 177, 95–101.DOI

Paper details

vlnc:W_bcwavmqx
Version
v2 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents