Unimodular Zeros of Selmer’s Harmonic Trinomials

Gerardo Barrera, Waldemar Barrera and Juan Pablo Navarrete

arXiv: 2610.22132v1 · PDF · 15 October 2026
Categories: math.CV, math.NT
Licence: CC BY 4.0 The site shows the text.

Abstract

Let n>k be positive integers and let \varepsilon,\delta\in\{-1,1\}. In this note, we classify the unimodular zeros of the harmonic trinomial H_{\varepsilon,\delta}:\mathbb{C}\to \mathbb{C} given by

H_{\varepsilon,\delta}(z)=z^n+\varepsilon\overline z^{\,k}+\delta,

where \overline{z} is the complex conjugate of z\in \mathbb{C}. Let g:=\gcd(n,k) be the greatest common divisor of n and k, and set n_0:=n/g and k_0:=k/g. For each choice of \varepsilon,\delta\in \{-1,1\}, we give a necessary and sufficient congruence involving n_0 and k_0 to guarantee that H_{\varepsilon,\delta} has unimodular zeros. More precisely, if the corresponding congruence holds, then H_{\varepsilon,\delta} has exactly 2g unimodular zeros, which are given explicitly; otherwise, it has no unimodular zeros. The proof uses a reduction to a trinomial equation and elementary divisibility arguments.

Items 9

Theorem

Theorem 1.1 (Unimodular zeros)

DNUS

The harmonic trinomial H_{\varepsilon,\delta} defined in (?) has unimodular zeros if and only if the corresponding condition below holds. The possible values of the gth powers of its unimodular zeros are listed in the third column.

The paper has a table here, which this page does not show.

Moreover, if the condition in the second column holds, then H_{\varepsilon,\delta} has exactly 2g distinct unimodular zeros.

Used by Proof

Proof of Theorem 1.1

Write w=z^g. By Lemma 1.3, it is enough to solve

w^{n_0}+\varepsilon w^{-k_0}+\delta=0, \quad w\in\mathbb{S}^1.

Suppose that (\varepsilon,\delta)=(1,1). The proofs of the other cases proceed similarly, using parts (b), (c), and (d) of Lemma 1.4 and Remark 1.5. Then (Proof) becomes w^{n_0}+w^{-k_0}+1=0. By part (a) of Lemma 1.4, we obtain \{w^{n_0},w^{-k_0}\}=\{\zeta,\zeta^2\}, which implies w^{3n_0}=w^{3k_0}=1. Since \gcd(n_0,k_0)=1, Bézout’s identity gives w^3=1. Note that w=1 does not satisfy w^{n_0}+w^{-k_0}+1=0. Hence, w\in\{\zeta,\zeta^2\}. Assume w=\zeta. Using the division Euclides algorithm, write n_0=3p_0+r_0 and k_0=3p'_0+r'_0 for some p_0,p'_0\in \mathbb{N}\cup\{0\} and r_0,r'_0\in \{0,1,2\}. We observe that

\begin{split} 0=w^{n_0}+w^{-k_0}+1&=\zeta^{n_0}+\zeta^{-k_0}+1= \zeta^{3p_0+r_0}+\zeta^{-3p'_0-r'_0}+1=\zeta^{r_0}+\zeta^{-r'_0}+1. \end{split}

Since |\zeta|=1, this implies r_0,r'_0\in \{1,2\}. If r_0=r'_0, the statement is complete. Assume without loss of generality that r_0=1 and r'_0=2. Note that

\begin{split} 0=\zeta+\zeta^{-2}+1=\zeta+\zeta+1, \end{split}

which yields a contradiction because |\zeta|=1.

The proof of the case w=\zeta^2 follows similarly.

Uses Lemma 1.3, Lemma 1.4, Remark 1.5

Corollary

Corollary 1.2

P7MS

For \varepsilon,\delta\in\{-1,1\}, define

\theta:=\begin{cases} 2\pi/3,&(\varepsilon,\delta)=(1,1),\\ \pi/3,&(\varepsilon,\delta)\ne(1,1). \end{cases}

Under the corresponding congruence in Theorem 1.1, the unimodular zeros of H_{\varepsilon,\delta} are given by

z_j^{\pm} =\exp\!\left(\mathsf{i}\frac{(\pm\theta+2\pi j)}{g}\right), \quad j=0,1,\ldots,g-1.

Proof of Corollary 1.2

In all cases covered by Theorem 1.1, the solutions of z^g=\mathrm{e}^{\pm\mathsf{i}\theta} are

z_j^{\pm} =\exp\left(\mathsf{i}\frac{(\pm\theta+2\pi j)}{g}\right), \quad j=0,\ldots,g-1.

Each sign gives g distinct solutions. The two sets are disjoint because \mathrm{e}^{\mathsf{i}\theta}\ne\mathrm{e}^{-\mathsf{i}\theta}. Consequently, there are exactly 2g unimodular zeros.

Uses Theorem 1.1

Lemma

Lemma 1.3 (Reduction lemma)

ESNZ

For z\in\mathbb{S}^1, set w:=z^g. Then

H_{\varepsilon,\delta}(z)=0 \quad\textrm{if and only if}\quad w^{n_0}+\varepsilon w^{-k_0}+\delta=0.

Every solution w\in\mathbb{S}^1 of the equation on the right-hand side of (1.3) has exactly g preimages in \mathbb{S}^1 under the map z\mapsto z^g.

Used by Proof

Proof of Lemma 1.3

Recall (?). For z\in\mathbb{S}^1, we note that z^n=(z^g)^{n_0}=w^{n_0} and \overline z^{\,k}=z^{-k}=(z^g)^{-k_0}=w^{-k_0}. Hence, H_{\varepsilon,\delta}(z)= w^{n_0}+\varepsilon w^{-k_0}+\delta, which proves (1.3).

Now, let w\in\mathbb{S}^1 be a solution of the equation on the right-hand side of (1.3). Write w=\mathrm{e}^{\mathsf{i} t} for some t\in\mathbb{R}. The solutions of z^g=w are given by

z_j:= \exp\left(\mathsf{i}\frac{(t+2\pi j)}{g}\right), \qquad j=0,\ldots,g-1.

These g points are all different and belong to \mathbb{S}^1. Therefore, every solution w\in\mathbb{S}^1 of the reduced equation on the right-hand side of (1.3) has exactly g preimages in \mathbb{S}^1.

Lemma

Lemma 1.4

HCGN

Let u,v\in\mathbb{S}^1 and \zeta:=\mathrm{e}^{2\pi\mathsf{i}/3}. Then the following statements hold.

\begin{aligned} \textrm{(a)}\quad u+v+1=0 &\quad\Longleftrightarrow\quad (u,v)=(\zeta,\zeta^2)\ \text{or}\ (\zeta^2,\zeta),\\ \textrm{(b)}\quad u+v-1=0 &\quad\Longleftrightarrow\quad (u,v)=(-\zeta,-\zeta^2)\ \text{or}\ (-\zeta^2,-\zeta),\\ \textrm{(c)}\quad u-v+1=0 &\quad\Longleftrightarrow\quad (u,v)=(\zeta,-\zeta^2)\ \text{or}\ (\zeta^2,-\zeta),\\ \textrm{(d)}\quad u-v-1=0 &\quad\Longleftrightarrow\quad (u,v)=(-\zeta,\zeta^2)\ \text{or}\ (-\zeta^2,\zeta). \end{aligned}

Used by Proof

Proof of Lemma 1.4

We prove (a); (b), (c), and (d) follow by applying (a) to (-u,-v), (u,-v), and (-u,v), respectively. Suppose u+v+1=0. Then v=-1-u, and

1=|v|^2=|1+u|^2=2+2\operatorname{Re}u.

Therefore, \operatorname{Re}u=-1/2. Write u=x+\mathsf{i} y with x,y\in\mathbb R and note that y=\pm\sqrt3/2. Hence, u=\zeta or u=\zeta^2. Since \zeta^3=1 and \zeta\ne1, we have 1+\zeta+\zeta^2=0. Therefore, v=-1-u is the other member of \{\zeta,\zeta^2\}. Conversely, both ordered pairs have unit-modulus entries and satisfy u+v+1=0.

Remark

Remark 1.5 (Orders of roots of unity)

L1DJ

Recall that \zeta:=\mathrm{e}^{2\pi\mathsf{i}/3}. Note that the cube roots of unity are 1,\zeta,\zeta^2 and have orders 1,3,3, respectively. Moreover, the sixth roots of unity are 1,-\zeta^2,\zeta,-1,\zeta^2,-\zeta, of orders 1,6,3,2,3,6, respectively. Indeed, for m\in\mathbb{Z}, (\zeta^2)^m=1 exactly when 3\mid2m, or, equivalently, when 3\mid m. The other orders follow in the same way from divisibility. Consequently, the only roots of unity of order six are -\zeta and -\zeta^2.

Used by Proof

Source and licence

The text on this page is from arXiv:2610.22132v1 by Gerardo Barrera, Waldemar Barrera and Juan Pablo Navarrete, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2610/2610.22132.jsonl.