Large values of cubic and quartic Dirichlet $L$-functions of prime moduli

Abstract.

In this paper, we apply the resonator method to exhibit large values at the central point of the families of cubic and quartic Dirichlet $L$-functions of prime moduli under the generalized Riemann hypothesis.

Mathematics Subject Classification (2010): 11M06, 11N56

Keywords: Dirichlet $L$-functions, large values, resonance method, cubic and quartic character sums

1. Introduction

Extreme values of $L$-functions have attracted much attention in the literature. In [22], K. Soundararajan introduced the resonance method to study large values of $L$-functions. This powerful method allowed him to show that for sufficiently large $X$,

$$\max_{\begin{subarray}{c}d\text{ fundamental discrimint }\\ X<|d|\leq 2X\end{subarray}}|L(\tfrac{1}{2},\chi_{d})|\geq\exp\left(\left(\frac{1}{\sqrt{5}}+o(1)\right)\sqrt{\frac{\log X}{\log_{2}X}}\right),$$

where $\chi_{d}=\left(\frac{d}{\cdot}\right)$ denotes the Kronecker symbol. Here and throughout the paper, we denote $\log_{j}$ the $j$-fold iterated logarithm.

The method of Soundararjan was subsequently utilized by C. Aistleitner [1], A. Bondarenko and K. Seip in [4], and by R. de la Bretèche and G. Tenenbaum in [7] to study large values of the Riemann zeta function $\zeta(s)$. Incorporating the long resonator method of Bondarenko–Seip in [4], P. Darbar and G. Maiti showed in [5] that under the generalized Riemann hypothesis (GRH), for sufficiently large $X$,

$$\max_{\begin{subarray}{c}d\text{ fundamental discrimint }\\ X<|d|\leq 2X\end{subarray}}|L(\tfrac{1}{2},\chi_{d})|\geq\exp\left(\left(\frac{1}{2}+o(1)\right)\sqrt{\frac{\log X\log_{3}X}{\log_{2}X}}\right).$$

Throughout the paper, we reserve the letter $p$ for a positive rational prime in $\mathbb{Z}$. It is shown in [9] by M. Fan, S. Hua and S. Xie using the resonance method that for sufficiently large $X$,

(1.1) $$\max_{\begin{subarray}{c}X<p\leq 2X\\ p\equiv 1\negthickspace\negthickspace\negthickspace\pmod{8}\end{subarray}}|L(\tfrac{1}{2},\chi_{p})|\geq\exp\left(\left(\sqrt{\frac{8}{45}}+o(1)\right)\sqrt{\frac{\log X}{\log_{2}X}}\right).$$

Based on the ideas in [5], the first-named author proved in [14] that under GRH, for sufficiently large $X$,

$$\begin{split}\max_{\begin{subarray}{c}X<p\leq 2X\end{subarray}}\Big|L(\tfrac{1}{2},\chi_{8p})\Big|\geq\exp\left(\left(\frac{1}{2}+o(1)\right)\sqrt{\frac{\log X\log_{3}X}{\log_{2}X}}\right).\end{split}$$

As noted in [5], one may adapt the approaches therein to show that the above inequality remains valid with $L(\frac{1}{2},\chi_{8p})$ replaced by $L(\frac{1}{2},\chi_{p})$ for a fundamental discriminant $p$, leading to an improvement of (1.1) under GRH.

For any $p$, let $\chi_{0}$ denote the principle Dirichlet character modulo $p$ (we omit writing out its dependence on $p$ as it will be clear from the context). We say a Dirichelt character $\chi$ modulo $p$ is a cubic (resp. quartic) character if $\chi^{3}=\chi_{0}$ (resp. $\chi^{4}=\chi_{0}$). It is the aim of this paper to extend the methods of Darbar and Maiti and study large values at the central point of the families of cubic and quartic Dirichlet $L$-functions of prime moduli under GRH.

Our main result is as follows.

Theorem 1.1.

With the notation as above and the truth of GRH, we have, for sufficiently large $X$,

(1.2) $$\max_{\begin{subarray}{c}X<p\leq 2X\\ \chi\bmod{p}\\ \chi^{3}=\chi_{0}\end{subarray}}\Big|L(\tfrac{1}{2},\chi)\Big|\geq\exp\left(\left(\frac{\sqrt{2}}{\pi}+o(1)\right)\frac{\log_{3}X}{\log_{2}X}\sqrt{\log X}\right),\\$$

and

(1.3) $$\max_{\begin{subarray}{c}X<p\leq 2X\\ \chi\bmod{p}\\ \chi^{4}=\chi_{0}\end{subarray}}\Big|L(\tfrac{1}{2},\chi)\Big|\geq\exp\left(\left(\frac{\sqrt{3/2}}{\pi}+o(1)\right)\frac{\log_{3}X}{\log_{2}X}\sqrt{\log X}\right).$$

2. Preliminaries

2.1. Cubic and quartic Dirichlet characters

Let $K$ be a number field and $\mathcal{O}_{K},U_{K}$ its ring of integers and group of units, respectively. We write $N(n)$ and $\mathrm{Tr}(n)$ for the norm and trace of any $n\in K$, respectively. Let $\chi$ denote a Hecke character of $K$ and we say that $\chi$ is of trivial infinite type if its component at infinite places of $K$ is trivial. We write $L(s,\chi)$ for the $L$-function associated to $\chi$ and $\zeta_{K}(s)$ the Dedekind zeta function of $K$. We shall reserve $\varpi$ for a prime number in $\mathcal{O}_{K}$ throughout the paper, by which we mean that the ideal $(\varpi)$ generated by $\varpi$ is a prime ideal.

Now suppose that $K$ is a number field of class number one. Let $\Lambda_{K}(n)$ denote the von Mangoldt function on $\mathcal{O}_{K}$ defined by

$$\Lambda_{K}(n)=\begin{cases}\log N(\varpi)\qquad&(n)=(\varpi^{k}),\text{$\varpi$ prime},k\geq 1,\\ 0\qquad&\text{otherwise}.\end{cases}$$

Let $n\in\mathbb{N}$ with $n\geq 2$ and $\mu_{n}(K):=\{\zeta\in K^{\times}:\zeta^{n}=1\}$ and suppose that $\mu_{n}(K)$ has $n$ elements. We can easily compute the discriminant of $x^{n}-1$ as

$$(-1)^{n(n-1)/2}R(x^{n}-1,nx^{n-1})=(-1)^{n(n-1)/2}n^{n}P=(-1)^{n(n+1)/2+1}n^{n}.$$

Here $R(f,g)$ is the resultant of the polynomials $f$ and $g$ and $P=(-1)^{n+1}$ is the product of the $n$-th roots of unity. Hence the discriminant of $x^{n}-1$ is divisible only by the primes dividing $n$ in $\mathcal{O}_{K}$. It follows that for any prime $\varpi\in\mathcal{O}_{K},(\varpi,n)=1$, we have a bijection

$$\zeta\mapsto\zeta\pmod{\varpi}:\mu_{n}(K)\rightarrow\mu_{n}(\mathcal{O}_{K}/\varpi):=\{\zeta\in(\mathcal{O}_{K}/\varpi)^{\times}:\zeta^{n}=1\}.$$

Thus $(\mathcal{O}_{K}/\varpi)^{\times}$ contains a subgroup of order $n$ which implies that $n|N(\varpi)-1$. For such $\varpi$, we define the $n$-th power residue symbol $\left(\frac{\cdot}{\varpi}\right)_{n,K}$ in $K$ such that $\left(\frac{a}{\varpi}\right)_{n,K}\equiv a^{(N(\varpi)-1)/n}\pmod{\varpi}$ with $\left(\frac{a}{\varpi}\right)_{n,K}\in\mu_{n}(K)$ for any $a\in\mathcal{O}_{K}$, $(a,\varpi)=1$ (see [17, Section 4.1]). When $\varpi|a$, we define $\left(\frac{a}{\varpi}\right)_{n,K}=0$. Then these symbols can be extended to any composite $c$ with $(N(c),n)=1$ multiplicatively. Further define $\left(\frac{\cdot}{c}\right)_{n,K}=1$ for $c\in U_{K}$.

In the remainder of this section, we set $K=\mathbb{Q}(\omega)$ with $\omega=\exp(2\pi i/3)$ or $K=\mathbb{Q}(i)$ unless otherwise specified. It is well-known that both fields have class number one and $\mathcal{O}_{K}=\mathbb{Z}[\omega]$, $\mathbb{Z}[i]$, respectively. We use $\delta_{K}$ and $D_{K}$ for the different and discriminant of $K$, respectively. In particular, we have $\delta_{\mathbb{Q}(\omega)}=\sqrt{-3}$, $\delta_{\mathbb{Q}(i)}=2i$, $D_{\mathbb{Q}(\omega)}=-3$, $D_{\mathbb{Q}(i)}=-4$. We shall reserve the symbol $\left(\frac{\cdot}{\cdot}\right)_{3}$ for the cubic residue symbol $\left(\frac{\cdot}{\cdot}\right)_{3,\mathbb{Q}(\omega)}$ and $\left(\frac{\cdot}{\cdot}\right)_{4}$ for the quartic residue symbol $\left(\frac{\cdot}{\cdot}\right)_{4,\mathbb{Q}(i)}$ in this paper.

Recall that every ideal in $\mathbb{Z}[\omega]$ co-prime to $3$ has a unique generator congruent to $1$ modulo $3$ (see [3, Proposition 8.1.4]) and every ideal in $\mathbb{Z}[i]$ co-prime to $2$ has a unique generator congruent to $1$ modulo $(1+i)^{3}$ (see the paragraph above Lemma 8.2.1 in [3])). These generators are called primary. An element $n=a+b\omega$ in $\mathbb{Z}[\omega]$ is congruent to $1\pmod{3}$ if and only if $a\equiv 1\pmod{3}$, and $b\equiv 0\pmod{3}$ (see the discussions before [15, Proposition 9.3.5]).

If $\varpi$ is a prime such that $N(\varpi)$ is a rational prime, then restricting $\chi_{3,\varpi}:=\left(\frac{\cdot}{\varpi}\right)_{3}$ or $\chi_{4,\varpi}:=\left(\frac{\cdot}{\varpi}\right)_{4}$ on rational integers gives rise to cubic or quartic Dirichlet characters, and we shall say that these Dirichlet characters are induced by $\chi_{3,\varpi}$ or $\chi_{4,\varpi}$. When no possible ambiguity exists, we write $\chi_{\varpi}$ for either $\chi_{3,\varpi}$ or $\chi_{4,\varpi}$. We have the following classification of primitive cubic and quartic Dirichlet characters of prime conductors from [12, Lemma 2.4].

Lemma 2.2.

The primitive cubic Dirichlet characters of prime conductor $p$ co-prime to $3$ are induced by $\chi_{3,\varpi}$ for some prime $\varpi\in\mathbb{Z}[\omega]\setminus\mathbb{Z}$ with $N(\varpi)=p$. The primitive quartic Dirichlet characters of prime conductor $p$ co-prime to $2$ such that their squares remain primitive are induced by $\chi_{4,\varpi}$ for some prime $\varpi\in\mathbb{Z}[i]\setminus\mathbb{Z}$ with $N(\varpi)=p$.

In particular, the above lemma implies that cubic Dirichlet characters of prime conductor $p$ exist if and only if $p\equiv 1\pmod{3}$, in which case there are two such characters induced by $\chi_{3,\varpi}$ or $\chi_{3,\overline{\varpi}}$, where $\varpi$ is a prime in $\mathbb{Z}[\omega]$ and $N(\varpi)=p$. In a similar vein, quartic Dirichlet characters of prime conductor $p$ exist precisely for $p\equiv 1\pmod{4}$. In that case, there are two such characters induced by $\chi_{4,\varpi}$ or $\chi_{4,\overline{\varpi}}$, where $\varpi$ is a prime in $\mathbb{Z}[i]$ with $N(\varpi)=p$.

2.3. The Gauss sums

As usual, set $e(x)=\exp(2\pi ix)$ and let $\chi$ be a Dirichlet character of modulus $n$. For any $r\in\mathbb{Z}$, the Gauss sum $\tau(r,\chi)$ is defined by

$$\tau(r,\chi)=\sum_{x\bmod{n}}\chi(x)e(rx).$$

We also write $\tau(\chi)$ for $\tau(1,\chi)$, for convenience.

For any number field $K$, we further define $\widetilde{e}_{K}(n)=e(\mathrm{Tr}(n/\delta_{K}))$ for any $n\in K$ and write $\widetilde{e}_{\omega}(z)$ and $\widetilde{e}_{i}(z)$ for $\widetilde{e}_{\mathbb{Q}(\omega)}(n)$ and $\widetilde{e}_{\mathbb{Q}(i)}(n)$, respectively. For any $n,r\in\mathbb{Z}[\omega],(n,3)=1$, set

$$g_{3}(r,n)=\sum_{x\bmod{n}}\left(\frac{x}{n}\right)_{3}\widetilde{e}_{\omega}\left(\frac{rx}{n}\right)$$

and, for any $n,r\in\mathbb{Z}[i],(n,2)=1$,

$$g_{4}(r,n)=\sum_{x\bmod{n}}\left(\frac{x}{n}\right)_{3}\widetilde{e}_{i}\left(\frac{rx}{n}\right).$$

We shall write $g_{j}(n)$ for $g_{j}(1,n)$ in what follows for $j=3,4$. We note the well-known formula from [19, p. 195]):

(2.1) $$|g_{j}(n)|= \begin{cases}\sqrt{N(n)},\qquad&\text{if $n$ is square-free},\\ 0,\qquad&\text{otherwise}.\end{cases}$$

Moreover, from [8], for $j=3$ and 4,

(2.2) $$g_{j}(rs,n)= \overline{\left(\frac{s}{n}\right)}_{j}g_{j}(r,n),\quad(s,n)=1.$$

If $\chi$ is induced by $\chi_{3,\varpi}$ for a primary $\varpi$, then (see [10, (2.7)])

(2.3) $$\tau(\chi)= \displaystyle\overline{\left(\frac{\sqrt{-3}}{\varpi}\right)}_{3}g_{3}(r,\varpi)=\overline{\left(\frac{\omega(1-\omega)}{\varpi}\right)}_{3}g_{3}(r,\varpi).$$

For a quartic character $\chi_{4,\varpi}$ with a primary $\varpi$, it is shown in [11, (2.11)] that

(2.4) $$\tau(\chi)=\begin{cases}\overline{\left(\frac{(-2i)^{3}}{n}\right)}_{4}g_{4}(n),\qquad&\text{if $\left(\frac{-1}{n}\right)_{4}=1$},\\ \\ i^{-1}\overline{\left(\frac{(-2i)^{3}}{n}\right)}_{4}g_{4}(n),\qquad&\text{if $\left(\frac{-1}{n}\right)_{4}=-1$ }.\end{cases}$$

A key step in the proof of Theorem 1.1 is the following bound established in [6, Theorem 4.4] and [13, Proposition 5.1] on a Dirichlet polynomial formed from cubic and quartic Gauss sums, respectively.

Proposition 2.4.

With the notation as above, let $j=3$, $4$ and $\psi$ be any ray class character modulo $9$ (resp. $16$) when $j=3$ (resp. $j=4$) and let $r$ be any primary element in $\mathcal{O}_{K}$. Then we have for real $Z>1$ and $\varepsilon>0$, we have

$$\begin{split}H_{Z}(r;\psi):=\sum_{\begin{subarray}{c}c\mathrm{\ primary}\\ (c,r)=1\\ N(c)\leq Z\end{subarray}}\frac{\Lambda_{K}(c)\psi(c)g_{K,j}(r,c)}{\sqrt{N(c)}}\ll&\begin{cases}\min(Z^{1+\varepsilon},N(r)^{1/10+\varepsilon}Z^{4/5+\varepsilon}),&j=3,\\ \\ \min(Z^{1+\varepsilon},N(r)^{1/16+\varepsilon}Z^{7/8+\varepsilon}),&j=4.\end{cases}\end{split}$$

2.5. The approximate functional equation

Let $\chi$ be any primitive Dirichlet character modulo $q$ and $\mathfrak{a}=0$ or $1$ be given by $\chi(-1)=(-1)^{\mathfrak{a}}$. We define

$$\Lambda(s,\chi)=\left(\frac{\pi}{q}\right)^{-(s+\mathfrak{a})/2}\Gamma\left(\frac{s+\mathfrak{a}}{2}\right)L(s,\chi).$$

Then $\Lambda(s,\chi)$ extends to an entire function on $\mathbb{C}$ when $\chi\neq\chi_{0}$ and satisfies the functional equation (see [16, Theorem 4.15])

$$\Lambda(1-s,\overline{\chi})=\frac{i^{\mathfrak{a}}q^{1/2}}{\tau(\chi)}\Lambda(s,\chi).$$

Let $G(s)$ be any even function which is holomorphic and bounded in the strip $-4<\Re(s)<4$ satisfying $G(0)=1$. From [16, Theorem 5.3], we have the following approximate functional equation for Dirichlet $L$-functions.

Proposition 2.6.

Suppose $\chi$ be a primitive Dirichlet character modulo $q$. Let $A$ and $B$ be positive real numbers with $AB=q$. Then we have

$$L\left(\frac{1}{2},\chi\right)=\sum_{m=1}^{\infty}\frac{\chi(m)}{m^{1/2}}V_{\mathfrak{a}}\left(\frac{m}{A}\right)+\epsilon(\chi)\sum_{m=1}^{\infty}\frac{\overline{\chi}(m)}{m^{1/2}}V_{\mathfrak{a}}\left(\frac{m}{B}\right),$$

where

$$\epsilon(\chi)=i^{-\mathfrak{a}}q^{-1/2}\tau(\chi),\quad V_{\mathfrak{a}}(x)=\frac{1}{2\pi i}\int\limits_{(2)}\frac{G(s)}{s}\gamma_{\mathfrak{a}}(s)x^{-s}\mathrm{d}s,\quad\gamma_{\mathfrak{a}}(s)=\pi^{-s/2}\frac{\Gamma\left(\tfrac{1/2+\mathfrak{a}+s}{2}\right)}{\Gamma\left(\tfrac{1/2+\mathfrak{a}}{2}\right)}.$$

If $\chi$ is a cubic character, then $\chi(-1)=\chi((-1)^{3})=1$. It follows that $\mathfrak{a}=0$ in this case. We now set $G(s)=1$ and for brevity write $V$ for $V_{0}$. Hence if $\chi$ is a cubic primitive Dirichlet character, then

(2.5) $$L\left(\tfrac{1}{2},\chi\right)=\sum_{m=1}^{\infty}\frac{\chi(m)}{m^{1/2}}V\left(\frac{m}{A}\right)+\epsilon(\chi)\sum_{m=1}^{\infty}\frac{\overline{\chi}(m)}{m^{1/2}}V\left(\frac{m}{B}\right).$$

For the quartic case, again set $G(s)=1$ and relabel $V_{\mathfrak{a}}$ by $V_{\pm}$ according to $\mathfrak{a}=0$ or $1$, arriving at

(2.6) $$L\left(\tfrac{1}{2},\chi\right)=\begin{cases}\displaystyle\sum_{m=1}^{\infty}\frac{\chi(m)}{m^{1/2}}V_{+}\left(\frac{m}{A}\right)+\epsilon(\chi)\sum_{m=1}^{\infty}\frac{\overline{\chi}(m)}{m^{1/2}}V_{+}\left(\frac{m}{B}\right),\quad\chi(-1)=1,\\ \\ \displaystyle\sum_{m=1}^{\infty}\frac{\chi(m)}{m^{1/2}}V_{-}\left(\frac{m}{A}\right)+\epsilon(\chi)\sum_{m=1}^{\infty}\frac{\overline{\chi}(m)}{m^{1/2}}V_{-}\left(\frac{m}{B}\right),\quad\chi(-1)=-1.\end{cases}$$

Note that for any real number $x>0$,

$$V(x)=\frac{1}{2\pi i}\int\limits\limits_{(2)}\left(\frac{1}{\pi}\right)^{s/2}\frac{\Gamma(s/2+1/4)}{\Gamma(1/4)}x^{-s}\frac{\mathrm{d}s}{s}.$$

Similar to what is shown in [21, Lemma 2.1], the function $V(x)$ is real-valued, smooth on $(0,+\infty)$ and for any $\varepsilon>0$,

(2.7) $$V\left(x\right)=1+O(x^{1/2-\varepsilon})\;\;\mbox{for}\;\;0<x<1\quad\mbox{and}\quad V^{(j)}\left(x\right)=O(e^{-x})\;\;\mbox{for}\;\;x>0,\;j\geq 0.$$

It is shown in [14] that $V(x)$ is positive for $x\geq 0$. The above properties are valid for $V_{\pm}(x)$ as well.

When $\chi=\chi_{\varpi}$ for some primary $\varpi\in\mathbb{Z}[\omega]\setminus\mathbb{Z}$ with $N(\varpi)=p$, then we deduce from (2.1), (2.3) and (2.5) that

(2.8) $$L\left(\tfrac{1}{2},\chi_{\varpi}\right)=\sum_{m=1}^{\infty}\frac{\chi_{\varpi}(m)}{m^{1/2}}V\left(\frac{m}{A}\right)+N(\varpi)^{-1/2}\sum_{m=1}^{\infty}\frac{\overline{\chi}_{\varpi}(\omega(1-\omega)m)g_{3}(\varpi)}{m^{1/2}}V\left(\frac{m}{B}\right).$$

If $\chi=\chi_{\varpi}$ for some primary prime $\varpi\in\mathbb{Z}[i]\setminus\mathbb{Z}$ with $N(\varpi)=p$, then (2.1), (2.4) and (2.6) similarly yield

(2.9) $$L\left(\tfrac{1}{2},\chi_{\varpi}\right)=\begin{cases}\displaystyle\sum_{m=1}^{\infty}\frac{\chi_{\varpi}(m)}{m^{1/2}}V_{+}\left(\frac{m}{A}\right)+N(\varpi)^{-1/2}\sum_{m=1}^{\infty}\frac{\overline{\chi}_{\varpi}((-2i)^{3}m)g_{4}(\varpi)}{m^{1/2}}V_{+}\left(\frac{m}{B}\right),\quad\chi_{\varpi}(-1)=1,\\ \\ \displaystyle\sum_{m=1}^{\infty}\frac{\chi_{\varpi}(m)}{m^{1/2}}V_{-}\left(\frac{m}{A}\right)-N(\varpi)^{-1/2}\sum_{m=1}^{\infty}\frac{\overline{\chi}_{\varpi}((-2i)^{3}m)g_{4}(\varpi)}{m^{1/2}}V_{-}\left(\frac{m}{B}\right),\quad\chi_{\varpi}(-1)=-1.\end{cases}$$

2.7. Smoothed character sums

We reserve $\psi_{m}$ for the Hecke characters in $K$ with $\psi_{m}((n))=\left(\frac{m}{n}\right)_{3}$ for $n\in\mathbb{Q}(\omega)$ coprime to $3$ or $\psi_{m}((n))=\left(\frac{m}{n}\right)_{4}$ for $n\in\mathbb{Q}(i)$ coprime to $2$. It is shown in [2, Section 2.1] and [11, Section 2.1] that $\psi_{m}$ is either a cubic Hecke character of trivial infinite type modulo $9m$ or a quartic Hecke character of trivial infinite type modulo $16m$. We define $\delta_{n=\text{cubic}}$ to be $1$ or $0$ depending on whether $n$ equals a cube or not, and $\delta_{n=\text{quartic}}$ similarly. Analogous to [20, Proposition 1], we need to bound smoothed sums of cubic and quartic characters in this paper. In the sequel, let $\Phi$ be a smooth, non-negative function compactly supported on $[1,2]$ and $\Phi(x)=1$ for $x\in[5/4,7/4]$. Denote the Mellin transform of $\Phi(x)$ by ${\widehat{\Phi}}(s)$ so that for any complex number $s$,

$${\widehat{\Phi}}(s)=\int\limits_{0}^{\infty}\Phi(x)x^{s}\frac{\mathrm{d}x}{x}.$$

We have the following result on the smoothed cubic and quartic Dirichlet character sums.

Lemma 2.8.

With the notation as above and assuming the truth of GRH, we have, for large $X$ and any positive integer $c\in\mathbb{Z}$,

(2.10) $$\sum_{\begin{subarray}{c}(p,3)=1\end{subarray}}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{3}=\chi_{0}\end{subarray}}(\log p)\chi(c)\Phi\Big(\frac{p}{X}\Big)=\displaystyle\delta_{c=\text{cubic}}{\widehat{\Phi}}(1)X+O(X^{1/2+\varepsilon}\log(c+2)),$$

and

(2.11) $$\sum_{(p,2)=1}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{4}=\chi_{0}\end{subarray}}\chi(c)\Phi\Big(\frac{p}{X}\Big)=\displaystyle\delta_{c=\text{quartic}}{\widehat{\Phi}}(1)X+O(X^{1/2+\varepsilon}\log(c+2)).$$

Here and after the asterisk on the sum over $\chi$ in (2.10) restricts the sum to primitive characters and $\chi_{0}$ denotes the principal character, while the asterisk on the sum over $\chi$ in (2.11) limits the sum to primitive characters $\chi$ such that $\chi^{2}$ remains primitive.

Proof.

As the proofs for both cases are similar, we shall only prove (2.10). Thus, in the remaining of the proof, we shall work exclusively over the number field $K=\mathbb{Q}(\omega)$ with all the related notations (e.g. primary) being referred to those related to this $K$. Let $CS$ denote the left-hand side of (2.10). Applying Lemma 2.2 gives

(2.12) $$\begin{split}CS=&\sum_{\begin{subarray}{c}\varpi\mathrm{\ primary}\\ \varpi\notin\mathbb{Z}\end{subarray}}(\log N(\varpi))\chi_{\varpi}(c)\Phi\left(\frac{N(\varpi)}{X}\right)\\ =&\sum_{\begin{subarray}{c}\varpi\mathrm{\ primary}\end{subarray}}(\log N(\varpi))\chi_{\varpi}(c)\Phi\left(\frac{N(\varpi)}{X}\right)+O\Big(\Big|\sum_{\begin{subarray}{c}p\in\mathbb{Z}\\ p\neq 3\end{subarray}}(\log N(p))\chi_{p}(c)\Phi\Big(\frac{N(p)}{X}\Big)\Big|\Big).\end{split}$$

For primes $p\in\mathbb{Z}$, $N(p)=p^{2}$. It follows that for any $\varepsilon>0$,

(2.13) $$\begin{split}\sum_{\begin{subarray}{c}p\in\mathbb{Z}\\ p\neq 3\end{subarray}}(\log N(p))\chi_{p}(c)\Phi\left(\frac{N(p)}{X}\right)\ll\sum_{\begin{subarray}{c}p\in\mathbb{Z}\end{subarray}}(\log p)\Phi\left(\frac{p^{2}}{X}\right)\ll X^{1/2+\varepsilon}.\end{split}$$

Thus, from (2.12) and (2.13),

(2.14) $$\begin{split}CS=&\sum_{\begin{subarray}{c}\varpi\mathrm{\ primary}\end{subarray}}(\log N(\varpi))\chi_{\varpi}(c)\Phi\left(\frac{N(\varpi)}{X}\right)+O\big(X^{1/2+\varepsilon}\big).\end{split}$$

As $\Phi$ has compact support, we get

(2.15) $$CS=\sum_{\begin{subarray}{c}\varpi\mathrm{\ primary}\end{subarray}}(\log N(\varpi))\chi_{\varpi}(c)\Phi\left(\frac{N(\varpi)}{X}\right)+O\big(X^{1/2+\varepsilon}\big)=\sum_{n\mathrm{\ primary}}\chi_{n}(c)\Lambda_{\omega}(n)\Phi\left(\frac{N(n)}{X}\right)+O\left(X^{1/2+\varepsilon}\right),$$

Via standard arguments, replacing $\log N(\varpi)$ by $\Lambda_{\omega}(n)$ incur an error that can be absorbed into the $O$-term above.

Now apply Mellin inversion renders

(2.16) $$\sum_{n\mathrm{\ primary}}\chi_{n}(c)\Lambda_{\mathbb{Q}(\omega)}(n)\Phi\left(\frac{N(n)}{X}\right)=-\frac{1}{2\pi i}\int\limits_{(2)}\frac{L^{\prime}(s,\psi_{c})}{L(s,\psi_{c})}\widehat{\Phi}(s)X^{s}\mathrm{d}s.$$

Note that integration by parts shows that $\widehat{\Phi}(s)$ satisfies the bound

(2.17) $$\widehat{\Phi}(s)\ll\min(1,|s|^{-1}(1+|s|)^{-E}),\;\mbox{for all}\;\Re(s)>0\;\mbox{and integers}E>0.$$

As discussed in Section 2.1, the Hecke character $\psi_{c}$ is a cubic Hecke character of trivial infinite type modulo $9c$. Suppose that it is induced by a primitive Hecke character $\widetilde{\psi}_{c}$ of trivial infinite type modulo $c^{\prime}$, then $c^{\prime}|9c$. Moreover,

$$L(s,\psi_{c})=L(s,\widetilde{\psi}_{c})\prod_{\begin{subarray}{c}(\varpi)\\ \varpi\mid 3c\end{subarray}}(1-\widetilde{\psi}_{c}(\varpi)N(\varpi)^{-s}).$$

It follows that

(2.18) $$-\frac{L^{\prime}(s,\psi_{c})}{L(s,\psi_{c})}=-\frac{L^{\prime}(s,\widetilde{\psi}_{c})}{L(s,\widetilde{\psi}_{c})}-\sum_{\begin{subarray}{c}(\varpi)\\ \varpi\mid 3c\end{subarray}}\frac{\widetilde{\psi}_{c}(\varpi)(\log N(\varpi))N(\varpi)^{-s}}{1-\widetilde{\psi}_{c}(\varpi)N(\varpi)^{-s}}.$$

For $\Re(s)\geq 1/2$, we have

(2.19) $$\sum_{\begin{subarray}{c}(\varpi)\\ \varpi\mid 3c\end{subarray}}\frac{\widetilde{\psi}_{c}(\varpi)(\log N(\varpi))N(\varpi)^{-s}}{1-\widetilde{\psi}_{c}(\varpi)N(\varpi)^{-s}}\ll\log(N(c)+2).$$

Also, it follows from [16, Theorem 5.17]) that under GRH, we have for $\Re(s)\geq 1/2+\varepsilon,|s-1|>1/100$,

(2.20) $$\frac{L^{\prime}(s,\widetilde{\psi}_{c})}{L(s,\widetilde{\psi}_{c})}\ll\log\big((N(c)+2)(1+|s|)\big).$$

We now evaluate the integral in (2.16) by shifting the line of integration to $\Re(s)=1/2+\varepsilon$. We encounter a pole at $s=1$ with residue $\widehat{\Phi}(1)X$ only if $c$ is a perfect cube. The integration on the line $\Re(s)=1/2+\varepsilon$ can be estimated as $O(X^{1/2+\varepsilon}\log(N(c)+2))$ by using (2.17) and (2.18)–(2.20). As $c\in\mathbb{Z}$, we have $N(c)=c^{2}$. This, together with (2.16), leads to (2.10) and completes the proof. ∎

3. Proof of the Theorem 1.1

3.1. Initial Treatments

Let $[x]$ denote the largest integer not exceeding $x$ for any real $x$ and $N\ll X$ be a large number to be fixed later. Suppose $a\in(1,\infty)$ and $\delta\in(0,1)$ are fixed with $1<a<1/\delta$. We define for $k=1,\ldots,[(\log_{2}N)^{\delta}]$ the set $\mathcal{P}_{k}$ to consist all primes $p$ such that

$$e^{k}\log N\log_{2}N<p\leq e^{k+1}\log N\log_{2}N.$$

Let $\mathcal{P}=\cup_{k}\mathcal{P}_{k}$. Thus $\mathcal{P}$ consists of all primes $p$ satisfying

$$e\log N\log_{2}N<p\leq e^{(\log_{2}N)^{\delta}}\log N\log_{2}N.$$

Denote by $\mathcal{R}_{k}$ the set of positive integers that have at least $\frac{a\log N}{k^{2}\log_{3}N}$ prime divisors in $\mathcal{P}_{k}$.

Define a multiplicative function $\psi$ supported on the set of square-free integers in the following manner. Let $\psi(p)=0$ for $p\notin\mathcal{P}$ and for any $p\in\mathcal{P}$,

(3.1) $$\psi(p)=\sqrt{\frac{\log N\log_{2}N}{\log_{3}N}}p^{-1/2}\left(\log p-\log(\log N\log_{2}N)\right)^{-1}.$$

Further set

(3.2) $$\mathcal{R}:=\rm{supp}(\psi)\setminus\displaystyle\bigcup_{k=1}^{[(\log_{2}N)^{\delta}]}\mathcal{R}_{k}.$$

The resonator for $L(\frac{1}{2},\chi)$ for any Dirichlet character $\chi$ is then defined to be the Dirichlet polynomial

$$R_{\chi}:=\sum_{m\in\mathcal{R}}\psi(m)\chi(m).$$

We quote the following result from [5, Lemma 3] on the size of $\mathcal{R}$, which clearly depends on $N$.

Lemma 3.2.

With the notation as above and $1<a<1/\delta$, we have $|\mathcal{R}|\leq N$ for $N$ large enough.

Now define

$$\mathcal{A}_{N}:=\frac{1}{\displaystyle\sum_{m\in\mathbb{N}}\psi(m)^{2}}\sum_{n\in\mathbb{N}}\frac{\psi{(n)}}{\sqrt{n}}\sum_{\begin{subarray}{c}l|n\end{subarray}}\psi(l)\sqrt{l}.$$

We further cite three results pertaining to $\mathcal{A}_{N}$ from [4, Lemmas 1–3].

Lemma 3.3.

With the notation as above, we have as $N\to\infty$,

$$\mathcal{A}_{N}\geq\exp\left((\delta+o(1))\sqrt{\frac{\log N\log_{3}N}{\log_{2}N}}\right).$$

Lemma 3.4.

With the notation as above, we have as $N\to\infty$,

$$\frac{1}{\displaystyle\sum_{m\in\mathbb{N}}\psi(m)^{2}}\sum_{\begin{subarray}{c}n\in\mathbb{N}\\ n\notin\mathcal{R}\end{subarray}}\frac{\psi{(n)}}{\sqrt{n}}\sum_{\begin{subarray}{c}l|n\end{subarray}}\psi(l)\sqrt{l}=o(\mathcal{A}_{N}).$$

Lemma 3.5.

With the notation as above, for any $\varepsilon>0$, we have as $N\to\infty$,

$$\frac{1}{\displaystyle\sum_{m\in\mathbb{N}}\psi(m)^{2}}\sum_{\begin{subarray}{c}n\in\mathcal{R}\end{subarray}}\frac{\psi{(n)}}{\sqrt{n}}\sum_{\begin{subarray}{c}l|n\\ l\leq n/N^{\varepsilon}\end{subarray}}\psi(l)\sqrt{l}=o(\mathcal{A}_{N}),$$

where the implicit constant only depends on $\varepsilon$.

3.6. Proof of (1.2)

Let $\Phi$ be the function described in Section 2.7. We set

$$\mathcal{S}_{1}:=\sum_{\begin{subarray}{c}(p,3)=1\end{subarray}}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\bmod{p}\\ \chi^{3}=\chi_{0}\end{subarray}}(\log p)L(\tfrac{1}{2},\chi)|R_{\chi}|^{2}\Phi\Big(\frac{p}{X}\Big)\quad\mbox{and}\quad\mathcal{S}_{2}:=\sum_{\begin{subarray}{c}(p,3)=1\end{subarray}}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\bmod{p}\\ \chi^{3}=\chi_{0}\end{subarray}}(\log p)|R_{\chi}|^{2}\Phi\Big(\frac{p}{X}\Big).$$

As $|R^{2}_{\chi}|\geq 0$, we then observe that

(3.3) $$\max_{\begin{subarray}{c}X<p\leq 2X\\ \chi\bmod{p}\\ \chi^{3}=\chi_{0}\end{subarray}}\Big|L(\tfrac{1}{2},\chi)\Big|\geq\frac{|\mathcal{S}_{1}|}{\mathcal{S}_{2}}.$$

Thus It remains to estimate $|\mathcal{S}_{1}|$ from below and $\mathcal{S}_{1}$ from above. Applying (2.5) and (2.8) leads to

(3.4) $$\begin{split}\mathcal{S}_{1}=&S_{1,1}+S_{1,2},\end{split}$$

where for some constant $A=A(X)>0$,

(3.5) $$\mathcal{S}_{1,1}=\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{l\geq 1}\frac{1}{\sqrt{l}}\sum_{\begin{subarray}{c}(p,3)=1\end{subarray}}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{3}=\chi_{0}\end{subarray}}(\log p)\chi(lm)\overline{\chi}(n)V\left(\frac{l}{A\sqrt{p}}\right)\Phi\Big(\frac{p}{X}\Big),$$

and

(3.6) $$\mathcal{S}_{1,2}=\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{l\geq 1}\frac{1}{\sqrt{l}}\sum_{\begin{subarray}{c}\varpi\mathrm{\ primary}\\ \varpi\notin\mathbb{Z}\end{subarray}}\frac{(\log N(\varpi))\chi_{\varpi}(m)\overline{\chi}_{\varpi}(\omega(1-\omega)ln)g_{3}(\varpi)}{\sqrt{N(\varpi)}}V\left(\frac{lA}{\sqrt{N(\varpi)}}\right)\Phi\left(\frac{N(\varpi)}{X}\right).$$

We treat $\mathcal{S}_{1,1}$ first. As $\chi$ is cubic, $\overline{\chi}(n)=\chi(n^{2})$. Applying Lemma 2.8 and partial summation, the inner double sum over $p$ and $\chi$ in $\mathcal{S}_{1,1}$ becomes

$$\begin{split}\sum_{\begin{subarray}{c}(p,3)=1\end{subarray}}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{3}=\chi_{0}\end{subarray}}&(\log p)\chi(lmn^{2})V\left(\frac{l}{A\sqrt{p}}\right)\Phi\Big(\frac{p}{X}\Big)\\ =&\int\limits^{2X}_{X}V\left(\frac{l}{A\sqrt{t}}\right)\mathrm{d}\left(t\delta_{lmn^{2}=\text{cubic}}{\widehat{\Phi}}(1)+O\left(t^{1/2+\varepsilon}\log(lmn^{2}+2)\right)\right),\\ =&{\widehat{\Phi}}(1)X\delta_{lmn^{2}=\text{cubic}}\int\limits^{2}_{1}V\left(\frac{l}{A\sqrt{Xt}}\right)\mathrm{d}t\\ &\hskip 28.45274pt+V\left(\frac{lA}{\sqrt{t}}\right)O\left(t^{1/2+\varepsilon}\log(lmn+2)\right)\Bigg|^{2X}_{X}-\int\limits^{2X}_{X}O\left(t^{1/2+\varepsilon}\log(lmn+2)\right)V^{\prime}\left(\frac{l}{A\sqrt{t}}\right)\frac{l}{2At^{3/2}}\mathrm{d}t\\ =:&{\widehat{\Phi}}(1)X\delta_{lmn^{2}=\text{cubic}}\int\limits^{2}_{1}V\left(\frac{l}{A\sqrt{Xt}}\right)\mathrm{d}t+R.\end{split}$$

Hence,

$$\mathcal{S}_{1,1}={\widehat{\Phi}}(1)X\sum_{m,n\in\mathcal{R}} \psi(m)\psi(n)\sum_{\begin{subarray}{c}lmn^{2}=\text{cubic}\end{subarray}}\frac{1}{\sqrt{l}}\int\limits_{1}^{2}V\left(\frac{l}{A\sqrt{Xt}}\right)dt+O\Bigg(\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{l\geq 1}\frac{R}{\sqrt{l}}\Bigg).$$

Mindful of the rapid decay of $V$ and $V^{\prime}$ given in (2.7), we may restrict the sum over $l$ to be $l\leq(A\sqrt{X})^{1+\varepsilon}$ for any $\varepsilon>0$ in the $O$-term above. It follows that

$$R\ll X^{1/2+\varepsilon}.$$

We deduce from the above, and Cauchy’s inequality, that

(3.7) $$\begin{split}\mathcal{S}_{1,1}=&{\widehat{\Phi}}(1)X\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{\begin{subarray}{c}lmn^{2}=\text{cubic}\end{subarray}}\frac{1}{\sqrt{l}}\int\limits_{1}^{2}V\left(\frac{l}{A\sqrt{Xt}}\right)\mathrm{d}t+O\Bigg(X^{1/2+\varepsilon}\sum_{\begin{subarray}{c}l\leq A^{1+\varepsilon}X^{1/2+\varepsilon}\end{subarray}}\frac{1}{\sqrt{l}}\left(\sum_{m\in\mathcal{R}}\psi(m)\right)^{2}\Bigg)\\ =&{\widehat{\Phi}}(1)X\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{\begin{subarray}{c}lmn^{2}=\text{cubic}\end{subarray}}\frac{1}{\sqrt{l}}\int\limits_{1}^{2}V\left(\frac{l}{A\sqrt{Xt}}\right)\mathrm{d}t+O\Bigg(X^{3/4+\varepsilon}A^{1/2+\varepsilon}|\mathcal{R}|\sum_{m\in\mathcal{R}}\psi^{2}(m)\Bigg).\end{split}$$

On to $\mathcal{S}_{1,2}$, we recast the inner sum over $\varpi$ as

(3.8) $$\begin{split}\sum_{\begin{subarray}{c}\varpi\mathrm{\ primary}\end{subarray}}&\frac{(\log N(\varpi))\chi_{\varpi}(m)\overline{\chi}_{\varpi}(\omega(1-\omega)ln)g_{3}(\varpi)}{\sqrt{N(\varpi)}}V\left(\frac{lA}{\sqrt{N(\varpi)}}\right)\Phi\left(\frac{N(\varpi)}{X}\right)\\ &+O\left(\Bigg|\sum_{\begin{subarray}{c}p\in\mathbb{Z}\end{subarray}}\frac{(\log N(p))\chi_{p}(m)\overline{\chi}_{p}(\omega(1-\omega)ln)g_{3}(p)}{\sqrt{N(p)}}V\left(\frac{lA}{\sqrt{N(p)}}\right)\Phi\left(\frac{N(p)}{X}\right)\Bigg|\right).\end{split}$$

Note that $N(p)=p^{2}$ for rational prime $p$. From (2.1) and the compact support of $\Phi$, we get

(3.9) $$\begin{split}\sum_{\begin{subarray}{c}p\in\mathbb{Z}\end{subarray}}\frac{(\log N(p))\chi_{p}(m)\overline{\chi}_{p}(\omega(1-\omega)ln)g_{3}(p)}{\sqrt{N(p)}}V\left(\frac{lA}{\sqrt{N(p)}}\right)\Phi\left(\frac{N(p)}{X}\right)\ll\sum_{\begin{subarray}{c}p\ll\sqrt{X}\end{subarray}}\log p\ll X^{1/2}.\end{split}$$

From this, (3.8) becomes

(3.10) $$\begin{split}\sum_{\begin{subarray}{c}c\mathrm{\ primary}\end{subarray}}&\frac{\Lambda_{\mathbb{Q}(\omega)}(c)\chi_{c}(m)\overline{\chi}_{c}(\omega(1-\omega)ln)g_{3}(c)}{\sqrt{N(c)}}V\left(\frac{lA}{\sqrt{N(c)}}\right)\Phi\left(\frac{N(c)}{X}\right)\\ &+O\Bigg(\Bigg|\sum_{\begin{subarray}{c}\varpi^{j}\\ \varpi\mathrm{\ primary},j\geq 2\end{subarray}}\frac{\Lambda_{\mathbb{Q}(\omega)}(\varpi^{j})\chi_{\varpi^{j}}(m)\overline{\chi}_{\varpi^{j}}(\omega(1-\omega)ln)g_{3}(\varpi^{j})}{\sqrt{N(\varpi^{j})}}V\left(\frac{lA}{\sqrt{N(\varpi^{j})}}\right)\Phi\left(\frac{N(\varpi^{j})}{X}\right)\Bigg|+X^{1/2}\Bigg).\end{split}$$

Arguing as in (3.9), we get, for any $\varepsilon>0$,

$$\begin{split}\sum_{\begin{subarray}{c}\varpi^{j}\\ \varpi\mathrm{\ primary},j\geq 2\end{subarray}}\frac{\Lambda_{\mathbb{Q}(\omega)}(\varpi^{j})\chi_{\varpi^{j}}(m)\overline{\chi}_{\varpi^{j}}(\omega(1-\omega)ln)g_{3}(\varpi^{j})}{\sqrt{N(\varpi^{j})}}V\left(\frac{lA}{\sqrt{N(\varpi^{j})}}\right)\Phi\left(\frac{N(\varpi^{j})}{X}\right)\ll X^{1/2+\varepsilon}.\end{split}$$

We deduce from the above and (3.10) that (3.8) is

(3.11) $$\begin{split}\sum_{\begin{subarray}{c}c\mathrm{\ primary}\end{subarray}}&\frac{\Lambda_{\mathbb{Q}(\omega)}(c)\chi_{c}(m)\overline{\chi}_{c}(\omega(1-\omega)ln)g_{3}(c)}{\sqrt{N(c)}}V\left(\frac{lA}{\sqrt{N(c)}}\right)\Phi\left(\frac{N(c)}{X}\right)+O\left(X^{1/2+\varepsilon}\right)\\ =&\sum_{\begin{subarray}{c}c\mathrm{\ primary}\\ (c,lmn)=1\end{subarray}}\frac{\Lambda_{\mathbb{Q}(\omega)}(c)\overline{\chi}_{c}(\omega(1-\omega))g_{3}(lnm^{2},c)}{\sqrt{N(c)}}V\left(\frac{lA}{\sqrt{N(c)}}\right)\Phi\left(\frac{N(c)}{X}\right)+O\left(X^{1/2+\varepsilon}\right),\end{split}$$

where we used the observation that $\chi(m)=\overline{\chi}(m^{2})$ as $\chi$ is cubic and (2.2).

Moreover, due to the rapid decay of $V$ given in (2.7), we may restrict the sum over $l$ in $S_{1,2}$ to $l\leq(\sqrt{X}/A)^{1+\varepsilon}$. It follows from this, (3.6), (3.11) and Cauchy’s inequality that

(3.12) $$\begin{split}\mathcal{S}_{1,2}=&\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{l\geq 1}\frac{1}{\sqrt{l}}\sum_{\begin{subarray}{c}c\mathrm{\ primary}\\ (c,lmn)=1\end{subarray}}\frac{\Lambda_{\mathbb{Q}(\omega)}(c)\overline{\chi}_{c}(\omega(1-\omega))g_{3}(lnm^{2},c)}{\sqrt{N(c)}}V\left(\frac{lA}{\sqrt{N(c)}}\right)\Phi\left(\frac{N(c)}{X}\right)\\ &\hskip 56.9055pt+O\left(X^{1/2+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{l\leq(\sqrt{X}/A)^{1+\varepsilon}}\frac{1}{\sqrt{l}}\right)\\ =&\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{l\geq 1}\frac{1}{\sqrt{l}}\sum_{\begin{subarray}{c}c\mathrm{\ primary}\\ (c,lmn)=1\end{subarray}}\frac{\Lambda_{\mathbb{Q}(\omega)}(c)\overline{\chi}_{c}(\omega(1-\omega))g_{3}(lnm^{2},c)}{\sqrt{N(c)}}V\left(\frac{lA}{\sqrt{N(c)}}\right)\Phi\left(\frac{N(c)}{X}\right)\\ &\hskip 56.9055pt+O\left(X^{3/4+\varepsilon}A^{-1/2+\varepsilon}|\mathcal{R}|\sum_{m\in\mathcal{R}}\psi^{2}(m)\right).\end{split}$$

Note that $\overline{\chi}_{c}(\omega(1-\omega))=\chi_{c}((\omega(1-\omega))^{2})$ and the Hecke character $\psi_{(\omega(1-\omega))^{2}}$ is a ray class character modulo $9$. This allows us to apply Proposition 2.4 and get

$$\begin{split}\sum_{\begin{subarray}{c}c\mathrm{\ primary}\\ (c,lmn)=1\end{subarray}}&\frac{\Lambda_{\mathbb{Q}(\omega)}(c)\overline{\chi}_{c}(\omega(1-\omega))g_{3}(lnm^{2},c)}{\sqrt{N(c)}}V\left(\frac{lA}{\sqrt{N(c)}}\right)\Phi\left(\frac{N(c)}{X}\right)\ll X^{\varepsilon}\int\limits^{2X}_{X}V\left(\frac{lA}{\sqrt{t}}\right)\Phi\left(\frac{t}{X}\right)\mathrm{d}(O(t^{4/5}N(lnm^{2})^{1/10}))\\ \ll&X^{\varepsilon}O(t^{4/5}N(lnm^{2})^{1/10})V\left(\frac{lA}{\sqrt{t}}\right)\Phi\left(\frac{t}{X}\right)\Bigg|^{2X}_{X}\\ &\hskip 56.9055pt-\int\limits^{2X}_{X}O(t^{4/5}N(lnm^{2})^{1/10})\Bigg(V^{\prime}\left(\frac{lA}{\sqrt{t}}\right)\Phi\left(\frac{t}{X}\right)\left(\frac{-lA}{2t^{3/2}}\right)+V\left(\frac{lA}{\sqrt{t}}\right)\Phi^{\prime}\left(\frac{t}{X}\right)\frac{1}{X}\Bigg)\mathrm{d}t.\end{split}$$

By the rapid decay of $V^{\prime}(x)$ given in (2.7), we may assume that $l\leq(\sqrt{t}/A)^{1+\varepsilon}$ for each fixed $t$ in the integral above. It follows from this and the above that

$$\begin{split}\sum_{\begin{subarray}{c}c\mathrm{\ primary}\\ (c,lmn)=1\end{subarray}}\frac{\Lambda_{\mathbb{Q}(\omega)}(c)\overline{\chi}_{c}(\omega(1-\omega))g_{3}(lnm^{2},c)}{\sqrt{N(c)}}V\left(\frac{lA}{\sqrt{N(c)}}\right)\Phi\left(\frac{N(c)}{X}\right)\ll X^{4/5+\varepsilon}(lnm^{2})^{1/5+\varepsilon}.\end{split}$$

Inserting the above into (3.12) and restricting the sum over $l$ to $l\leq(\sqrt{X}/A)^{1+\varepsilon}$, we arrive at

(3.13) $$\begin{split}\mathcal{S}_{1,2}\ll&\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{l\leq(\sqrt{X}/A)^{1+\varepsilon}}\frac{X^{4/5+\varepsilon}}{\sqrt{l}}(lnm^{2})^{1/5+\varepsilon}+O\left(X^{3/4+\varepsilon}A^{-1/2+\varepsilon}|\mathcal{R}|\sum_{m\in\mathcal{R}}\psi^{2}(m)\right)\\ \ll&X^{23/20+\varepsilon}A^{-7/10+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{2})^{1/5+\varepsilon}+O\left(X^{3/4+\varepsilon}A^{-1/2+\varepsilon}|\mathcal{R}|\sum_{m\in\mathcal{R}}\psi^{2}(m)\right).\end{split}$$

We now deduce from (3.4), (3.12), (3.13) and Lemma 3.2,

$$\begin{split}|\mathcal{S}_{1}|\geq{\widehat{\Phi}}(1)X&\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{\begin{subarray}{c}lmn^{2}=\text{cubic}\end{subarray}}\frac{1}{\sqrt{l}}\int\limits_{1}^{2}V\left(\frac{l}{A\sqrt{Xt}}\right)\mathrm{d}t\\ &+O\Bigg(X^{3/4+\varepsilon}A^{1/2+\varepsilon}N\sum_{m\in\mathcal{R}}\psi^{2}(m)+X^{23/20+\varepsilon}A^{-7/10+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{2})^{1/5+\varepsilon}\Bigg).\end{split}$$

As both $\psi$ and $V$ are both non-negative, we may further minorize the leading term in (3.7) by keeping only the terms with $lm=n$ and discarding the rest. Thus we obtain

(3.14) $$\begin{split}|\mathcal{S}_{1}|\geq{\widehat{\Phi}}(1)X\sum_{n\in\mathcal{R}}\frac{\psi(n)}{\sqrt{n}}&\sum_{\begin{subarray}{c}m\mid n\end{subarray}}\psi(m)\sqrt{m}\int\limits_{1}^{2}V\left(\frac{n}{mA\sqrt{Xt}}\right)\mathrm{d}t\\ &+O\Bigg(X^{3/4+\varepsilon}A^{1/2+\varepsilon}N\sum_{m\in\mathcal{R}}\psi^{2}(m)+X^{23/20+\varepsilon}A^{-7/10+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{2})^{1/5+\varepsilon}\Bigg).\end{split}$$

As $N\ll X$, it follows that if $m\geq n/N^{\varepsilon}$ and $A\geq 1$, $0<n/(mA\sqrt{Xt})<1$ for $1\leq t\leq 2$. Applying the estimate $V(x)=1+O\left(x^{1/2-\varepsilon}\right)$ in (2.7), we to deduce from (3.14) that

(3.15) $$\begin{split}|\mathcal{S}_{1}|\geq{\widehat{\Phi}}(1)X\sum_{n\in\mathcal{R}}\frac{\psi(n)}{\sqrt{n}}&\sum_{\begin{subarray}{c}m\mid n\\ m\geq n/N^{\varepsilon}\end{subarray}}\psi(m)\sqrt{m}+O\Bigg(X^{3/4+\varepsilon}A^{-1/2+\varepsilon}\sum_{n\in\mathcal{R}}\psi(n)\sum_{\begin{subarray}{c}m\mid n\\ m\geq n/N^{\varepsilon}\end{subarray}}\psi(m)\Bigg)\\ &+O\Bigg(X^{3/4+\varepsilon}A^{1/2+\varepsilon}N\sum_{m\in\mathcal{R}}\psi^{2}(m)+X^{23/20+\varepsilon}A^{-7/10+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{2})^{1/5+\varepsilon}\Bigg).\end{split}$$

We observe from (3.1) that we have $0<\psi(p)\leq 1$ for $p\in\mathcal{P}$. It follows from this that

$$\sum_{n\in\mathcal{R}}\psi(n)\sum_{\begin{subarray}{c}m\mid n\\ m\geq n/N^{\varepsilon}\end{subarray}}\psi(m)\ll\sum_{n\in\mathcal{R}}\psi(n)\sum_{\begin{subarray}{c}m\mid n\end{subarray}}\psi(m)=\sum_{m\in\mathcal{R}}\psi(m)\sum_{\begin{subarray}{c}n\in\mathcal{R}\\ m\mid n\end{subarray}}\psi(n)\ll|\mathcal{R}|\sum_{m\in\mathcal{R}}\psi^{2}(m).$$

As $|\mathcal{R}|\leq N$ from Lemma 3.2, we deduce from the above and (3.15) that

(3.16) $$\begin{split}|\mathcal{S}_{1}|\geq&{\widehat{\Phi}}(1)X\sum_{n\in\mathcal{R}}\frac{\psi(n)}{\sqrt{n}}\sum_{\begin{subarray}{c}m\mid n\\ m\geq n/N^{\varepsilon}\end{subarray}}\psi(m)\sqrt{m}\\ &+O\Bigg(X^{3/4+\varepsilon}A^{1/2+\varepsilon}N\sum_{m\in\mathcal{R}}\psi^{2}(m)+X^{23/20+\varepsilon}A^{-7/10+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{2})^{1/5+\varepsilon}\Bigg).\end{split}$$

Note that we have by [18, (1.26)] that

(3.17) $$\begin{split}\sum^{[(\log_{2}N)^{\delta}]}_{k=1}\frac{1}{k}=\log[(\log_{2}N)^{\delta}]+\gamma_{0}+O([(\log_{2}N)^{\delta}]^{-1}),\end{split}$$

where $\gamma_{0}$ is the Euler constant. We now deduce, from (3.17) and the definition of $\mathcal{R}$ given in (3.2) that for any $n\in\mathcal{R}$,

$$\begin{split}n\ll\prod_{1\leq k\leq[(\log_{2}N)^{\delta}]}(e^{k+1}\log N\log_{2}N)^{a\log N/(k^{2}\log_{3}N)}\ll N^{a\delta+\varepsilon}e^{a\pi^{2}\log N\log_{2}N/(6\log_{3}N)},\end{split}$$

where the last estimation above follows by noting that $\sum^{\infty}_{k=1}1/k^{2}=\zeta(2)=\pi^{2}/6$. As $a\delta<1$, we see from the above that for any $n\in\mathcal{R}$, we have

$$\begin{split}n\ll N^{a\pi^{2}\log_{2}N/(6\log_{3}N)+\varepsilon}.\end{split}$$

Using the above, together with Cauchy’s inequality, we infer

(3.18) $$\begin{split}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{2})^{1/5+\varepsilon}\ll&N^{a\pi^{2}\log_{2}N/(10\log_{3}N)+\varepsilon}|\mathcal{R}|\sum_{m\in\mathcal{R}}\psi(m)^{2}\\ \ll&N^{a\pi^{2}\log_{2}N/(10\log_{3}N)+\varepsilon}\sum_{m\in\mathcal{R}}\psi(m)^{2},\end{split}$$

again using the bound $|\mathcal{R}|\leq N$, furnished by Lemma 3.2.

It follows from (3.18) that

(3.19) $$\begin{split}X^{3/4+\varepsilon}A^{1/2+\varepsilon}&N\sum_{m\in\mathcal{R}}\psi^{2}(m)+X^{23/20+\varepsilon}A^{-7/10+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{2})^{1/5+\varepsilon}\\ \ll&\Big(X^{3/4+\varepsilon}A^{1/2+\varepsilon}N+X^{23/20+\varepsilon}A^{-7/10+\varepsilon}N^{a\pi^{2}\log_{2}N/(10\log_{3}N)+\varepsilon}\Big)\sum_{m\in\mathcal{R}}\psi(m)^{2}.\end{split}$$

The optimal choice of $A$ is

$$\begin{split}A=X^{1/3}N^{a\pi^{2}\log_{2}N/(12\log_{3}N)}.\end{split}$$

and renders that the right-hand side of (3.19) is

(3.20) $$\ll X^{11/12+\varepsilon}N^{a\pi^{2}\log_{2}N/(24\log_{3}N)+\varepsilon}\sum_{m\in\mathcal{R}}\psi(m)^{2}.$$

We now apply (3.16), (3.20) together with Lemmas 3.3–3.5 to see that

(3.21) $$\begin{split}|\mathcal{S}_{1}|\geq\left(1+o(1)\right){\widehat{\Phi}}(1)&X\exp\left((\delta+o(1))\sqrt{\frac{\log N\log_{3}N}{\log_{2}N}}\right)\left(\sum_{\begin{subarray}{c}m\in\mathcal{R}\end{subarray}}\psi(m)^{2}\right)\\ &+O\left(X^{11/12+\varepsilon}N^{a\pi^{2}\log_{2}N/(24\log_{3}N)+\varepsilon}\sum_{m\in\mathcal{R}}\psi(m)^{2}\right).\end{split}$$

We now set the value of $N$ to be such that

(3.22) $$\begin{split}N^{a\pi^{2}\log_{2}N/(24\log_{3}N)}=X^{1/12-5\varepsilon}.\end{split}$$

Recall that we have $1<a<1/\delta$ so that the above choice of $N$ implies that

(3.23) $$\begin{split}\frac{\log N\log_{2}N}{\log_{3}N}=\frac{24}{a\pi^{2}}\Big(\frac{1}{12}-5\varepsilon\Big)\log X>\frac{24\delta}{\pi^{2}}\Big(\frac{1}{12}-5\varepsilon\Big)\log X.\end{split}$$

We deduce from (3.21)–(3.23) that

(3.24) $$\begin{split}|\mathcal{S}_{1}|\geq&\left(1+o(1)\right){\widehat{\Phi}}(1)X\exp\left(\left(\delta\sqrt{\frac{24\delta}{\pi^{2}}\Big(\frac{1}{12}-5\varepsilon\Big)}+o(1)\right)\frac{\log_{3}X}{\log_{2}X}\sqrt{\log X}\right)\left(\sum_{\begin{subarray}{c}m\in\mathcal{R}\end{subarray}}\psi(m)^{2}\right).\end{split}$$

Next, we proceed to establish an upper bound of $\mathcal{S}_{2}$. Again, we apply Lemma 2.8 and use arguments similar to those leading to (3.14). Hence,

$$\mathcal{S}_{2}= \sum_{\begin{subarray}{c}(p,3)=1\end{subarray}}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{3}=\chi_{0}\end{subarray}}(\log p)|R_{\chi}|^{2}\Phi\Big(\frac{p}{X}\Big)=\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{\begin{subarray}{c}(p,3)=1\end{subarray}}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{3}=\chi_{0}\end{subarray}}(\log p)\chi(mn^{2})\Phi\Big(\frac{p}{X}\Big)$$

$$= {\widehat{\Phi}}(1)X\sum_{\begin{subarray}{c}m,n\in\mathcal{R}\\ mn^{2}=\text{cubic}\end{subarray}}\psi(m)\psi(n)+O\Bigg(X^{1/2+\varepsilon}\sum_{\begin{subarray}{c}m,n\in\mathcal{R}\end{subarray}}\psi(m)\psi(n)\log(mn^{2}+2)\Bigg)$$

$$= {\widehat{\Phi}}(1)X\sum_{\begin{subarray}{c}n\in\mathcal{R}\end{subarray}}\psi(n)^{2}+O\left(X^{1/2+\varepsilon}|\mathcal{R}|\sum_{\begin{subarray}{c}m\in\mathcal{R}\end{subarray}}\psi(m)^{2}\right),$$

where the last equality above follows by noting that $mn^{2}$ is a cube only if $m=n$ since $m$ and $n$ are both square-free.

Again, using $|\mathcal{R}|\leq N=X^{o(1)}$, we deduce from the above that

(3.25) $$\mathcal{S}_{2}\leq\left(1+o(1)\right){\widehat{\Phi}}(1)X\left(\sum_{\begin{subarray}{c}m\in\mathcal{R}\end{subarray}}\psi(m)^{2}\right).$$

Hence, for sufficiently large $X$ and arbitrary small $\varepsilon>0$, we deduce from (3.3), (3.24) and (3.25) that

$$\max_{\begin{subarray}{c}X<p\leq 2X\\ \chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{3}=\chi_{0}\end{subarray}}\Big|L(\tfrac{1}{2},\chi)\Big| \geq\frac{|\mathcal{S}_{1}|}{\mathcal{S}_{2}}\geq\exp\left(\left(\delta\sqrt{\frac{24\delta}{\pi^{2}}\Big(\frac{1}{12}-5\varepsilon\Big)}+o(1)\right)\frac{\log_{3}X}{\log_{2}X}\sqrt{\log X}\right).$$

With $\delta\rightarrow 1^{-}$, we obtain the desired estimation for the first expression given in (1.2).

3.7. Proof of (1.3)

As the proof is similar to that given in the previous section, we shall be brief here. Keeping the notation in the previous section, set

$$\mathcal{T}_{1}:=\sum_{\begin{subarray}{c}(p,2)=1\end{subarray}}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{4}=\chi_{0}\end{subarray}}(\log p)L(\tfrac{1}{2},\chi)|R_{\chi}|^{2}\Phi\Big(\frac{p}{X}\Big)\quad\mbox{and}\quad\mathcal{T}_{2}:=\sum_{\begin{subarray}{c}(p,2)=1\end{subarray}}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{4}=\chi_{0}\end{subarray}}(\log p)|R_{\chi}|^{2}\Phi\Big(\frac{p}{X}\Big).$$

Then we have

(3.26) $$\max_{\begin{subarray}{c}X<p\leq 2X\\ \chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{4}=\chi_{0}\end{subarray}}\Big|L(\tfrac{1}{2},\chi)\Big|\geq\frac{|\mathcal{T}_{1}|}{\mathcal{T}_{2}}.$$

We first apply (2.6), (2.9) and the expression $(1\pm\chi(-1))/2$ to detect those characters satisfying $\chi(-1)=\pm 1$ to see that

(3.27) $$\begin{split}\mathcal{T}_{1}=&T_{1,1,+}+T_{1,1,-}+T_{1,2,+}+T_{1,2,-},\end{split}$$

where for some constant $B=B(X)>0$,

$$\begin{split}\mathcal{T}_{1,1,\pm}=&\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{l\geq 1}\frac{1}{\sqrt{l}}\sum_{\begin{subarray}{c}(p,2)=1\end{subarray}}\ \sideset{}{{}^{*}}{\sum}_{\begin{subarray}{c}\chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{4}=\chi_{0}\end{subarray}}(\log p)\frac{1\pm\chi(-1)}{2}\chi(lm)\overline{\chi}(n)V_{\pm}\left(\frac{l}{B\sqrt{p}}\right)\Phi\Big(\frac{p}{X}\Big),\quad\mbox{and}\\ \mathcal{T}_{1,2,\pm}=&\pm\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{l\geq 1}\frac{1}{\sqrt{l}}\\ &\hskip 28.45274pt\times\sum_{\begin{subarray}{c}\varpi\mathrm{\ primary}\\ \varpi\notin\mathbb{Z}\end{subarray}}\frac{1\pm\chi_{\varpi}(-1)}{2}\cdot\frac{(\log N(\varpi))\chi_{\varpi}(m)\overline{\chi}_{\varpi}((-2i)^{3}ln)g_{4}(\varpi)}{\sqrt{N(\varpi)}}V_{\pm}\left(\frac{lB}{\sqrt{N(\varpi)}}\right)\Phi\left(\frac{N(\varpi)}{X}\right).\end{split}$$

Similar to (3.7),

(3.28) $$\begin{split}\mathcal{T}_{1,1,+}+\mathcal{T}_{1,1,-}=\frac{1}{2}{\widehat{\Phi}}(1)X\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)&\sum_{\begin{subarray}{c}lmn^{3}=\text{quartic}\end{subarray}}\frac{1}{\sqrt{l}}\int\limits_{1}^{2}\left(V_{+}\left(\frac{l}{B\sqrt{Xt}}\right)+V_{-}\left(\frac{l}{B\sqrt{Xt}}\right)\right)\mathrm{d}t\\ &+O\Bigg(X^{3/4+\varepsilon}B^{1/2+\varepsilon}|\mathcal{R}|\sum_{m\in\mathcal{R}}\psi^{2}(m)\Bigg).\end{split}$$

We apply Proposition 2.4, analogous to (3.13), getting

(3.29) $$\begin{split}T_{1,2,+}+T_{1,2,-}\ll&\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{l\leq(\sqrt{X}/B)^{1+\varepsilon}}\frac{X^{7/8+\varepsilon}}{\sqrt{l}}(lnm^{3})^{1/8+\varepsilon}+O\left(X^{3/4+\varepsilon}B^{-1/2+\varepsilon}|\mathcal{R}|\sum_{m\in\mathcal{R}}\psi^{2}(m)\right)\\ \ll&X^{19/16+\varepsilon}B^{-5/8+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{3})^{1/8+\varepsilon}+O\left(X^{3/4+\varepsilon}B^{-1/2+\varepsilon}|\mathcal{R}|\sum_{m\in\mathcal{R}}\psi^{2}(m)\right).\end{split}$$

From (3.27)–(3.29) and Lemma 3.2, we decude that

(3.30) $$\begin{split}|\mathcal{T}_{1}|\geq\frac{1}{2}{\widehat{\Phi}}(1)X&\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)\sum_{\begin{subarray}{c}lmn^{3}=\text{quartic}\end{subarray}}\frac{1}{\sqrt{l}}\int\limits_{1}^{2}\left(V_{+}\left(\frac{l}{B\sqrt{Xt}}\right)+V_{-}\left(\frac{l}{B\sqrt{Xt}}\right)\right)\mathrm{d}t\\ &+O\Bigg(X^{3/4+\varepsilon}B^{1/2+\varepsilon}N\sum_{m\in\mathcal{R}}\psi^{2}(m)+X^{19/16+\varepsilon}B^{-5/8+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{3})^{1/8+\varepsilon}\Bigg).\end{split}$$

The positivity of $\psi$ and $V$ again allows us to keep only the terms with $lm=n$ in the main term of (3.30). Hence, similar to (3.16),

(3.31) $$\begin{split}|\mathcal{T}_{1}|\geq{\widehat{\Phi}}(1)X&\sum_{n\in\mathcal{R}}\frac{\psi(n)}{\sqrt{n}}\sum_{\begin{subarray}{c}m\mid n\\ m\geq n/N^{\varepsilon}\end{subarray}}\psi(m)\sqrt{m}\\ &+O\Bigg(X^{3/4+\varepsilon}B^{1/2+\varepsilon}N\sum_{m\in\mathcal{R}}\psi^{2}(m)+X^{19/16+\varepsilon}B^{-5/8+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{3})^{1/8+\varepsilon}\Bigg).\end{split}$$

Similar to (3.18),

(3.32) $$\begin{split}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{3})^{1/8+\varepsilon}\ll&N^{a\pi^{2}\log_{2}N/(12\log_{3}N)+\varepsilon}\sum_{m\in\mathcal{R}}\psi(m)^{2}.\end{split}$$

It follows from (3.32) that

(3.33) $$\begin{split}X^{3/4+\varepsilon}B^{1/2+\varepsilon}&N\sum_{m\in\mathcal{R}}\psi^{2}(m)+X^{19/16+\varepsilon}B^{-5/8+\varepsilon}\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{3})^{1/8+\varepsilon}\\ \ll&\Big(X^{3/4+\varepsilon}B^{1/2+\varepsilon}N+X^{19/16+\varepsilon}B^{-5/8+\varepsilon}N^{a\pi^{2}\log_{2}N/(12\log_{3}N)+\varepsilon}\Big)\sum_{m\in\mathcal{R}}\psi(m)^{2}.\end{split}$$

We now optimize our choice of $B$ as

$$\begin{split}B=X^{7/18}N^{2a\pi^{2}\log_{2}N/(27\log_{3}N)}.\end{split}$$

When inserted into (3.33), this yields

(3.34) $$\begin{split}X^{3/4+\varepsilon}A^{1/2+\varepsilon}N\sum_{m\in\mathcal{R}}\psi^{2}(m)+X^{19/16+\varepsilon}B^{-5/8+\varepsilon}&\sum_{m,n\in\mathcal{R}}\psi(m)\psi(n)(nm^{3})^{1/8+\varepsilon}\\ \ll&X^{17/18+\varepsilon}N^{a\pi^{2}\log_{2}N/(27\log_{3}N)+\varepsilon}\sum_{m\in\mathcal{R}}\psi(m)^{2}.\end{split}$$

We now apply (3.31), (3.34) together with Lemmas 3.3–3.5 to see that

(3.35) $$\begin{split}|\mathcal{T}_{1}|\geq\left(1+o(1)\right){\widehat{\Phi}}(1)X\exp&\left((\delta+o(1))\sqrt{\frac{\log N\log_{3}N}{\log_{2}N}}\right)\left(\sum_{\begin{subarray}{c}m\in\mathcal{R}\end{subarray}}\psi(m)^{2}\right)\\ &+O\left(X^{17/18+\varepsilon}N^{a\pi^{2}\log_{2}N/(27\log_{3}N)+\varepsilon}\sum_{m\in\mathcal{R}}\psi(m)^{2}\right).\end{split}$$

We now choose the value of $N$ satisfying

(3.36) $$\begin{split}N^{a\pi^{2}\log_{2}N/(27\log_{3}N)}=X^{1/18-5\varepsilon}.\end{split}$$

Recall that we have $1<a<1/\delta$ so that the above implies that

(3.37) $$\begin{split}\frac{\log N\log_{2}N}{\log_{3}N}=\frac{27}{a\pi^{2}}\Big(\frac{1}{18}-5\varepsilon\Big)\log X>\frac{27\delta}{\pi^{2}}\Big(\frac{1}{18}-5\varepsilon\Big)\log X.\end{split}$$

We deduce from (3.35)–(3.37) that

(3.38) $$\begin{split}|\mathcal{T}_{1}|\geq&\left(1+o(1)\right){\widehat{\Phi}}(1)X\exp\left(\left(\delta\sqrt{\frac{27\delta}{\pi^{2}}\Big(\frac{1}{18}-5\varepsilon\Big)}+o(1)\right)\frac{\log_{3}X}{\log_{2}X}\sqrt{\log X}\right)\left(\sum_{\begin{subarray}{c}m\in\mathcal{R}\end{subarray}}\psi(m)^{2}\right).\end{split}$$

Next, similar to (3.25), we have

$$\mathcal{T}_{2}\leq\left(1+o(1)\right){\widehat{\Phi}}(1)X\left(\sum_{\begin{subarray}{c}m\in\mathcal{R}\end{subarray}}\psi(m)^{2}\right).$$

Hence, for sufficiently large $X$ and arbitrary small $\varepsilon>0$, (3.26), (3.38) and the above render

$$\max_{\begin{subarray}{c}X<p\leq 2X\\ \chi\negthickspace\negthickspace\negthickspace\pmod{p}\\ \chi^{4}=\chi_{0}\end{subarray}}\Big|L(\tfrac{1}{2},\chi)\Big| \geq\frac{|\mathcal{T}_{1}|}{\mathcal{T}_{2}}\geq\exp\left(\left(\delta\sqrt{\frac{27\delta}{\pi^{2}}\Big(\frac{1}{18}-5\varepsilon\Big)}+o(1)\right)\frac{\log_{3}X}{\log_{2}X}\sqrt{\log X}\right).$$

By taking $\delta\rightarrow 1^{-}$, we arrive at the desired bound in (1.3).

Acknowledgments. P. G. is supported in part by NSFC grant 12471003 and L. Z. by the FRG Grant PS71536 at the University of New South Wales.