数学杂志  2019, Vol. 39 Issue (3): 379-386   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
王晓瑛
曹艳梅
关于短区间的并集中D.H.Lehmer问题的一个推广
王晓瑛, 曹艳梅    
西北大学数学学院, 陕西 西安 710127
摘要:本文研究了短区间的并集中的D.H.Lehmer问题.利用不完全Kloosterman和的均值定理,给出了D.H.Lehmer问题的渐近公式,从而推广了短区间的D.H.Lehmer问题.
关键词D.H.Lehmer问题    不完全Kloosterman和    短区间    
ON A GENERALIZATION OF THE D. H. LEMHER PROBLEM IN UNIONS OF SHORT INTERVALS
WANG Xiao-ying, CAO Yan-mei    
School of Mathematics, Northwest University, Xi'an 710127, China
Abstract: In this paper, the D. H. Lemher problem in unions of short intervals is studied. By using the mean value theorem for incomplete Kloosterman sums, we give an asymptotic formula for D. H. Lemher problem, which generalize the D. H. Lemher problem in short intervals.
Keywords: D. H. Lemher problem     incomplete Kloosterman sum     short interval    
1 引言

设整数$ q>2 $.对任意与$ q $互素的整数$ a $, 存在唯一的整数$ b $满足$ 1\leq b\leq q $以及$ ab\equiv 1 (\bmod\ q) $. D. H. Lehmer [1]建议研究$ a $$ b $的奇偶性不同的情形.当$ q = p $为奇素数时, 张文鹏[2]证明了

$ \begin{eqnarray*} \mathop{\mathop{\mathop{\sum\limits_{a = 1}^p} \mathop{\sum\limits_{b = 1}^p}}_{ab\equiv 1(\bmod p)}}_{2\nmid a+b}1 = \frac{1}{2}p+O\left(p^{\frac{1}{2}} \log^2 p\right). \end{eqnarray*} $

随后在文献[3, 4]中, 张文鹏还得到了渐近公式

$ \begin{eqnarray*} \mathop{\mathop{\mathop{\sum\limits_{a = 1}^q}_{(a, q) = 1} \mathop{\sum\limits_{b = 1}^q}_{(b, q) = 1}}_{ab\equiv 1(\bmod q)}}_{2\nmid a+b}1 = \frac{1}{2}\phi(q)+O\left(q^{\frac{1}{2}} d^2(q) \log^2 q\right), \end{eqnarray*} $

其中$ \phi(q) $为Euler函数, $ d(q) $是除数函数.

$ k $为非负整数.张文鹏[5]进一步证明了

$ \begin{eqnarray*} \mathop{\mathop{\mathop{\sum\limits_{a = 1}^q}_{(a, q) = 1} \mathop{\sum\limits_{b = 1}^q}_{(b, q) = 1}}_{ab\equiv 1(\bmod q)}}_{2\nmid a+b}\left(a-b\right)^{2k} = \frac{1}{(2k+1)(2k+2)}\phi(q)q^{2k}+O\left(4^k q^{2k+\frac{1}{2}}d^2(q)\log^2 q\right). \end{eqnarray*} $

此外设$ 0\leq x, y \leq 1 $, 文献[5]中还得到了

$ \begin{eqnarray*} \mathop{\mathop{\mathop{\sum\limits_{a\leq xq}}_{(a, q) = 1}\mathop{\sum\limits_{b\leq yq}}_{(b, q) = 1}}_{ab\equiv 1 (\bmod q)}}_{2\nmid a+b}1 = \frac{1}{2}xy\phi(q)+O\left(q^{\frac{1}{2}}d^2(q)\log^2q\right). \end{eqnarray*} $

设实数$ l, \delta $满足$ l\geq0 $$ 0<\delta\leq 1 $.王晓瑛与赵秋红在文献[6]中给出了渐近公式

$ \begin{eqnarray*} &&\mathop{\mathop{\mathop{\sum\limits_{a\leq xq}}_{(a, q) = 1}\mathop{\sum\limits_{b\leq yq}}_{(b, q) = 1}}_{ab\equiv 1 (\bmod q)}}_{2\nmid a+b}|a-b|^l = \frac{2x^{l+2}}{(l+1)(l+2)}\varphi(q)q^l+O\left(q^{l+\frac{1}{2}}d(q)\log q(\sigma_{-\frac{1}{2}}(q)+\sigma_{-\frac{3}{2}}(q)\log q)\right), \\ &&\mathop{\mathop{\mathop{\mathop{\sum\limits_{a\leq xq}}_{(a, q) = 1}\mathop{\sum\limits_{b\leq yq}}_{(b, q) = 1}}_{ab\equiv 1 (\bmod q)}}_{2\nmid a+b}}_{|a-b|<\delta q}|a-b|^l = \varphi(q)q^l\left(\frac{x\delta^{l+1}}{l+1}-\frac{\delta^{l+2}}{l+2}\right) +O\left(q^{l+\frac{1}{2}}d(q)\log q(\sigma_{-\frac{1}{2}}(q)+\sigma_{-\frac{3}{2}}(q)\log q)\right). \end{eqnarray*} $

$ q, c, n $为整数, 满足$ n\geq 2 $, $ q\geq 3 $以及$ (n, q) = (c, q) = 1 $.设$ 0<\delta_1, \delta_2\leq 1 $.陆亚明与易媛[7]给出了D. H. Lemher问题的推广

$ \begin{eqnarray*} \mathop{\mathop{\mathop{\sum\limits_{a\leq \delta_1q}}_{(a, q) = 1}\mathop{\sum\limits_{b\leq \delta_2q}}_{(b, q) = 1}}_{ab\equiv c(\bmod q)}}_{n\nmid a+b}1 = \left(1-\frac{1}{n}\right)\delta_1\delta_2\phi(q)+O\left(q^\frac{1}{2}d^6(q)\log^2q\right). \end{eqnarray*} $

本文进一步考虑D. H. Lemher问题在短区间的并集上的推广.主要结论如下.

定理1.1   设$ p $是奇素数, $ H>0 $, $ K>0 $, 并设$ I_1^{(j)} $, $ I_2^{(j)} $$ (0, p) $的子区间, $ 1\leq j\leq J $, 满足$ |I_1^{(j)}| = H $, $ |I_2^{(j)}| = K $, 以及$ I_1^{(j)}\bigcap I_1^{(k)} = \emptyset $, 当$ j\neq k $时.设$ c $, $ n $为整数, 满足$ n\geq 2 $以及$ (n, p) = (c, p) = 1 $.则有

$ \begin{eqnarray*} \sum\limits_{j = 1}^{J}\mathop{\mathop{{\mathop{{\sum}}_{x \in I_1^{(j)}}\mathop{{\sum}}_{y \in I_2^{(j)}}}}_{xy\equiv c (\bmod p)}}_{n\nmid x+y}1 = \left(1-\frac{1}{n}\right)\frac{JHK}{p}+O\left(J^{\frac{1}{2}}p^{\frac{1}{2}}\log p\log H\right). \end{eqnarray*} $

推论1.1   当$ J\gg \frac{p^{3}\log^{4}p}{H^{2}K^{2}} $时, 存在$ j\in \left\{ 1,2,\cdots ,J \right\} $, 使得方程$ x \in I_1^{(j)}, $ $ y \in I_2^{(j)}, $ $ xy\equiv c (\bmod\ p), $ $ n\nmid(x+y) $有解.

2 Kloosterman和的估计

$ p>2 $为素数, $ m $$ n $为任意整数.经典的Kloosterman和的定义为

$ \begin{eqnarray*} K(m, n;p) = \sum\limits_{a = 1}^{p-1}e\left(\frac{ma+n\overline{a}}{p}\right), \end{eqnarray*} $

其中$ e(y) = \hbox{e}^{2\pi iy} $, $ \overline{a} $表示$ a $关于模$ p $的逆, 满足$ 1\leq \overline{a}\leq p-1 $以及$ a\overline{a}\equiv 1 (\bmod\ p) $.由文献[8]可得著名的上界估计$ K(m, n;p)\ll p^{\frac{1}{2}}(m, n, p)^{\frac{1}{2}}. $

Browning和Haynes [9]给出了短区间的并集上的Kloosterman和的某种估计式.

引理2.1   设$ p $是奇素数, $ H $为正整数, $ I_{1}, \cdots , I_{J} $$ (0, p) $的互不相交的子区间, 且对任意$ j $满足$ H/2<|I_{j}|\leq H $.设整数$ l $$ p $互素, 则有

$ \begin{eqnarray*} \sum\limits_{j = 1}^{J}\left|\sum\limits_{n\in I_{j}}e\left(\frac{l\overline{n}}{p}\right)\right|^{2}\leq 2^{12}p\log^{2}H. \end{eqnarray*} $

为了证明本文的定理, 需要进一步考虑短区间的并集上的Kloosterman和的估计.

引理2.2   设$ p $是奇素数, $ H, n, s, l $为整数, 满足$ 1\leq H\leq p $, $ n\geq 2 $以及$ (n, p) = (l, p) = 1 $.则有

$ \begin{eqnarray*} \sum\limits_{r = 1}^{p}\left|\mathop{\sum\limits_{m = r+1}^{r+H}}_{m\not\equiv 0 (\bmod p)}e\left(\frac{sm}{n}+\frac{l\overline{m}}{p}\right)\right|^2\ll pH. \end{eqnarray*} $

  由剩余系的性质可得

$ \begin{eqnarray*} &&\sum\limits_{r = 1}^{p}\left|\mathop{\sum\limits_{m = r+1}^{r+H}}_{m\not\equiv 0(\bmod p)} e\left(\frac{sm}{n}+\frac{l\overline{m}}{p}\right)\right|^2 = \sum\limits_{r = 1}^{p}\left|\mathop{\sum\limits_{h = 1}^{H}}_{r\not\equiv -h(\bmod p)} e\left(\frac{s(r+h)}{n}+\frac{l\overline{r+h}}{p}\right)\right|^2\\ & = &\sum\limits_{r = 1}^{p}\mathop{\sum\limits_{h_{1} = 1}^{H}}_{r\not\equiv -h_{1}(\bmod p)} e\left(\frac{s(r+h_{1})}{n}+\frac{l\overline{r+h_{1}}}{p}\right) \mathop{\sum\limits_{h_{2} = 1}^{H}}_{r\not\equiv -h_{2}(\bmod p)} e\left(-\frac{s(r+h_{2})}{n}-\frac{l\overline{r+h_{2}}}{p}\right)\\ & = &\sum\limits_{h_{1} = 1}^{H}e\left(\frac{sh_{1}}{n}\right) \sum\limits_{h_{2} = 1}^{H}e\left(-\frac{sh_{2}}{n}\right)\mathop{\sum\limits_{r = 1}^{p}}_{r\not\equiv -h_{1}, -h_{2}(\bmod p)} e\left(\frac{l(\overline{r+h_{1}}-\overline{r+h_{2}})}{p}\right)\\ & = &\sum\limits_{h_{1} = 1}^{H}e\left(\frac{sh_{1}}{n}\right) \sum\limits_{h_{2} = 1}^{H}e\left(-\frac{sh_{2}}{n}\right) \mathop{\sum\limits_{x = 1}^{p-1}\sum\limits_{y = 1}^{p-1}}_{x-y\equiv h_{1}-h_{2}(\bmod p)} e\left(\frac{l(\overline{x}-\overline{y})}{p}\right)\\ & = &\frac{1}{p}\sum\limits_{h_{1} = 1}^{H}e\left(\frac{sh_{1}}{n}\right) \sum\limits_{h_{2} = 1}^{H}e\left(-\frac{sh_{2}}{n}\right) \sum\limits_{x = 1}^{p-1}\sum\limits_{y = 1}^{p-1} e\left(\frac{l(\overline{x}-\overline{y})}{p}\right) \sum\limits_{a = 1}^{p}e\left(\frac{a(x-y)+a(h_{2}-h_{1})}{p}\right)\\ & = &\frac{1}{p}\sum\limits_{h_{1} = 1}^{H}e\left(\frac{sh_{1}}{n}\right) \sum\limits_{h_{2} = 1}^{H}e\left(-\frac{sh_{2}}{n}\right) \sum\limits_{a = 1}^{p}e\left(\frac{a(h_{2}-h_{1})}{p}\right) \sum\limits_{x = 1}^{p-1}e\left(\frac{ax+l\overline{x}}{p}\right) \sum\limits_{y = 1}^{p-1}e\left(-\frac{ay+l\overline{y}}{p}\right)\\ & = &\frac{1}{p}\sum\limits_{a = 1}^{p}\left|\sum\limits_{h = 1}^{H}e\left(\frac{sh}{n}-\frac{ah}{p}\right)\right|^2 \left|\sum\limits_{x = 1}^{p-1}e\left(\frac{ax+l\overline{x}}{p}\right)\right|^{2}. \end{eqnarray*} $

再由Kloosterman和的经典估计, 可得

$ \begin{eqnarray*} &&\sum\limits_{r = 1}^{p}\left|\mathop{\sum\limits_{m = r+1}^{r+H}}_{m\not\equiv 0(\bmod p)} e\left(\frac{sm}{n}+\frac{l\overline{m}}{p}\right)\right|^2 \ll\sum\limits_{a = 1}^{p}\left|\sum\limits_{h = 1}^{H}e\left(\frac{sh}{n}-\frac{ah}{p}\right)\right|^2\\ & = &\sum\limits_{h_{1} = 1}^{H}e\left(\frac{sh_{1}}{n}\right) \sum\limits_{h_{2} = 1}^{H}e\left(-\frac{sh_{2}}{n}\right) \sum\limits_{a = 1}^{p}e\left(\frac{a(h_{2}-h_{1})}{p}\right)\\ & = &pH. \end{eqnarray*} $

引理2.3   设$ p $是奇素数, $ I_1, \cdots, I_J\subseteq (0, p) $是互不相交的子区间, 且对任意的$ j $满足$ \frac{H}{2}<|I_j|\leq H $, 其中$ H\leq p $为正整数.设$ n, s, l $为整数, 满足$ n\geq 2 $以及$ (n, p) = (l, p) = 1 $.则有

$ \begin{eqnarray*} \sum\limits_{j = 1}^J\left|\sum\limits_{r\in I_j}e\left(\frac{sr}{n}+\frac{l\overline{r}}{p}\right)\right|^2\ll p\log^{2}H. \end{eqnarray*} $

  利用引理2.2以及文献[9]中的方法, 不难证明引理2.3.为了完整起见, 在此给出详细的证明.

$ R_{j} $为集合$ I_{j} $中的最小正整数, 并假设$ R_{1}<R_{2}<\cdot\cdot\cdot <R_{j} $.显然有$ R_{j+1}-R_{j}\geq \frac{H}{2} $.定义

$ \begin{eqnarray*} S(r, H) = \mathop{\sum\limits_{m = r+1}^{r+H}}_{m\not\equiv 0 (\bmod p)}e\left(\frac{sm}{n}+\frac{l\overline{m}}{p}\right). \end{eqnarray*} $

容易证明

$ \begin{eqnarray} \sum\limits_{j = 1}^J\left|\sum\limits_{r\in I_j}e\left(\frac{sr}{n}+\frac{l\overline{r}}{p}\right)\right|^2\leq\sum\limits_{j = 1}^J \max\limits_{1\leq h\leq H}\left|S(R_{j}, h)\right|^{2}. \end{eqnarray} $ (2.1)

$ 1\leq h\leq H $$ R_{j}-H<r\leq R_{j} $, 可得

$ \begin{eqnarray*} \left|S(R_{j}, h)\right| = \left|S(r, R_{j}-r+h)-S(r, R_{j}-r)\right| \leq 2\max\limits_{1\leq k\leq 2H}\left|S(r, k)\right|. \end{eqnarray*} $

因此有

$ \begin{eqnarray*} \left|S(R_{j}, h)\right| &\leq&\frac{2}{H}\sum\limits_{R_{j}-H<r\leq R_{j}}\max\limits_{1\leq k\leq2H}\left|S(r, k)\right|. \end{eqnarray*} $

再由柯西不等式, 可得

$ \begin{eqnarray*} \left|S(R_{j}, h)\right|^{2} &\leq&\frac{4}{H^{2}}\left(\sum\limits_{R_{j}-H<r\leq R_{j}}\max\limits_{1\leq k\leq2H}\left|S(r, k)\right|\right)^{2}\\ &\leq&\frac{4}{H}\sum\limits_{R_{j}-H<r\leq R_{j}}\max\limits_{1\leq k\leq2H}\left|S(r, k)\right|^{2}. \end{eqnarray*} $

$ h $取最大值, 并对$ j $求和, 有

$ \begin{eqnarray*} &&\sum\limits_{j = 1}^J\max\limits_{1\leq h\leq H}\left|S(R_{j}, h)\right|^{2} \leq\frac{4}{H}\sum\limits_{j = 1}^J\sum\limits_{R_{j}-H<r\leq R_{j}}\max\limits_{1\leq k\leq2H}\left|S(r, k)\right|^{2}\\ & = &\frac{4}{H}\mathop{\sum\limits_{j = 1}^J}_{2|j}\sum\limits_{R_{j}-H<r\leq R_{j}}\max\limits_{1\leq k\leq 2H}\left|S(r, k)\right|^{2} +\frac{4}{H}\mathop{\sum\limits_{j = 1}^J}_{2\nmid j}\sum\limits_{R_{j}-H<r\leq R_{j}}\max\limits_{1\leq k\leq2H}\left|S(r, k)\right|^{2}. \end{eqnarray*} $

注意到$ R_{j+1}-R_{j}\geq \frac{H}{2} $, 易证

$ \begin{eqnarray} \sum\limits_{j = 1}^J\left|\sum\limits_{r\in I_j}e\left(\frac{sr}{n}+\frac{l\overline{r}}{p}\right)\right|^2 \leq\frac{8}{H}\sum\limits_{r = 1}^{p}\max\limits_{1\leq k\leq 2H}\left|S(r, k)\right|^{2}. \end{eqnarray} $ (2.2)

取正整数$ t $, 满足$ 2H\leq2^{t}\leq4H $.此时显然有$ t+1\leq4\log H $.对任意的$ 1\leq r\leq p $, 选择合适的正整数$ k = k(r)\leq2H $, 使得$ \mathop {\max }\limits_{1 \le h \le 2H} \left|S(r, h)\right| = \left|S(r, k)\right| $.记$ k = \sum\limits_{d \in D} {{2^{t - d}}} $, 其中$ D $$ [0, t] $之间整数的某个集合, 因此

$ \begin{eqnarray*} S(r, k) = \sum\limits_{d\in D}S(r+v_{r, d}2^{t-d}, 2^{t-d}), \end{eqnarray*} $

其中$ {v_{r, d}} = \sum\limits_{\begin{array}{*{20}{c}} {e \in D}\\ {e \in d} \end{array}} {{2^{d - e}} < {2^d}}. $

利用柯西不等式, 有

$ \begin{eqnarray*} &&\max\limits_{1\leq h\leq2H}\left|S(r, h)\right|^{2} = \left|S(r, k)\right|^{2} = \left(\sum\limits_{d\in D}S(r+v_{r, d}2^{t-d}, 2^{t-d})\right)^{2}\\ &\leq&|D|\sum\limits_{d\in D}\left|S(r+v_{r, d}2^{t-d}, 2^{t-d})\right|^{2}\\ &\leq&(t+1)\sum\limits_{0\leq d\leq t}\sum\limits_{0\leq v< 2^{d}}\left|S(r+v2^{t-d}, 2^{t-d})\right|^{2}. \end{eqnarray*} $

上式两边对$ r $求和, 并结合引理2.2, 有

$ \begin{eqnarray} \sum\limits_{r = 1}^{p}\max\limits_{1\leq h\leq2H}\left|S(r, h)\right|^{2} &\leq&(t+1)\sum\limits_{0\leq d\leq t}\sum\limits_{0\leq v<2^{d}}\sum\limits_{r = 1}^{p}\left|S(r+v2^{t-d}, 2^{t-d})\right|^{2}\\ &\ll&(t+1)\sum\limits_{0\leq d\leq t}2^{d}(p2^{t-d})\ll(t+1)^{2}p2^{t}\\ &\ll&pH\log^{2}H. \end{eqnarray} $ (2.3)
$ \begin{eqnarray} \end{eqnarray} $ (2.4)

结合(2.1)–(2.3)式, 立即可得

$ \begin{eqnarray*} \sum\limits_{j = 1}^J\left|\sum\limits_{r \in I_j}e\left(\frac{sr}{n}+\frac{l\overline{r}}{p}\right)\right|^2 &\ll&p\log^{2}H. \end{eqnarray*} $
3 定理1.1的证明

易证

$ \begin{eqnarray*} \sum\limits_{j = 1}^{J}\mathop{\mathop{{\mathop{{\sum}}_{x \in I_1^{(j)}}\mathop{{\sum}}_{y \in I_2^{(j)}}}}_{xy\equiv c (\bmod p)}}_{n\nmid x+y}1 = \sum\limits_{j = 1}^{J}\mathop{{\mathop{{\sum}}_{x \in I_1^{(j)}}\mathop{{\sum}}_{y \in I_2^{(j)}}}}_{xy\equiv c (\bmod p)}1-\sum\limits_{j = 1}^{J}\mathop{\mathop{{\mathop{{\sum}}_{x \in I_1^{(j)}}\mathop{{\sum}}_{y \in I_2^{(j)}}}}_{xy\equiv c (\bmod p)}}_{n\mid x+y}1. \end{eqnarray*} $

由三角恒等式, 有

$ \begin{eqnarray*} &&\sum\limits_{j = 1}^{J}\mathop{{\mathop{{\sum}}_{x \in I_1^{(j)}}\mathop{{\sum}}_{y \in I_2^{(j)}}}}_{xy\equiv c (\bmod p)}1 = \frac{1}{p}\sum\limits_{j = 1}^{J}\sum\limits_{x \in I_1^{(j)}}\sum\limits_{y \in I_2^{(j)}}\sum\limits_{|l|\leq\frac{p-1}{2}}e\left(\frac{l(c\overline{x}-y)}{p}\right)\\ & = &\frac{1}{p}\sum\limits_{|l|\leq\frac{p-1}{2}}\sum\limits_{j = 1}^{J}\sum\limits_{y \in I_2^{(j)}}e\left(-\frac{ly}{p}\right) \sum\limits_{x \in I_1^{(j)}}e\left(\frac{lc\overline{x}}{p}\right)\\ & = &\frac{JHK}{p}+\frac{1}{p}\mathop{\sum\limits_{|l|\leq\frac{p-1}{2}}}_{l\neq 0}\sum\limits_{j = 1}^{J}\sum\limits_{y \in I_2^{(j)}}e\left(-\frac{ly}{p}\right) \sum\limits_{x \in I_1^{(j)}}e\left(\frac{lc\overline{x}}{p}\right). \end{eqnarray*} $

再由柯西不等式以及引理2.1可得

$ \begin{eqnarray*} &&\frac{1}{p}\mathop{\sum\limits_{|l|\leq\frac{p-1}{2}}}_{l\neq 0}\sum\limits_{j = 1}^{J}\sum\limits_{y \in I_2^{(j)}}e\left(-\frac{ly}{p}\right) \sum\limits_{x \in I_1^{(j)}}e\left(\frac{lc\overline{x}}{p}\right)\\ &\ll&\frac{1}{p}\mathop{\sum\limits_{|l|\leq\frac{p-1}{2}}}_{l\neq 0}\sum\limits_{j = 1}^{J}\left|\sum\limits_{y \in I_2^{(j)}}e\left(-\frac{ly}{p}\right)\right| \cdot\left|\sum\limits_{x \in I_1^{(j)}}e\left(\frac{lc\overline{x}}{p}\right)\right|\\ &\ll&\mathop{\sum\limits_{|l|\leq\frac{p-1}{2}}}_{l\neq 0}\frac{1}{|l|}\sum\limits_{j = 1}^{J}\left|\sum\limits_{x \in I_1^{(j)}}e\left(\frac{lc\overline{x}}{p}\right)\right| \ll\mathop{\sum\limits_{|l|\leq\frac{p-1}{2}}}_{l\neq 0}\frac{1}{|l|}\cdot J^{\frac{1}{2}}\left(\sum\limits_{j = 1}^{J}\left|\sum\limits_{x \in I_1^{(j)}}e\left(\frac{lc\overline{x}}{p}\right)\right|^2\right)^{\frac{1}{2}}\\ &\ll& J^{\frac{1}{2}}p^{\frac{1}{2}}\log p\log H. \end{eqnarray*} $

因此

$ \begin{eqnarray} \sum\limits_{j = 1}^{J}\mathop{\mathop{{\mathop{{\sum}}_{x \in I_1^{(j)}}\mathop{{\sum}}_{y \in I_2^{(j)}}}}_{xy\equiv c (\bmod p)}}_{n\nmid x+y}1 = \frac{JHK}{p}-\sum\limits_{j = 1}^{J}\mathop{\mathop{{\mathop{{\sum}}_{x \in I_1^{(j)}}\mathop{{\sum}}_{y \in I_2^{(j)}}}}_{xy\equiv c (\bmod p)}}_{n\mid x+y}1+O\left(J^{\frac{1}{2}}p^{\frac{1}{2}}\log p\log H\right). \end{eqnarray} $ (3.1)

另一方面, 由三角恒等式有

$ \begin{eqnarray} &&\sum\limits_{j = 1}^{J}\mathop{\mathop{{\mathop{{\sum}}_{x \in I_1^{(j)}}\mathop{{\sum}}_{y \in I_2^{(j)}}}}_{xy\equiv c (\bmod p)}}_{n\mid x+y}1 = \frac{1}{p}\sum\limits_{j = 1}^{J}\mathop{\sum\limits_{x \in I_1^{(j)}}\sum\limits_{y \in I_2^{(j)}}}_{n\mid x+y}\sum\limits_{l = 1}^{p}e\left(\frac{l(c\overline{x}-y)}{p}\right)\\ & = &\frac{1}{p}\sum\limits_{j = 1}^{J}\mathop{\sum\limits_{x \in I_1^{(j)}}\sum\limits_{y \in I_2^{(j)}}}_{n\mid x+y}1+\frac{1}{p}\sum\limits_{l = 1}^{p-1}\sum\limits_{j = 1}^{J}\mathop{\sum\limits_{x \in I_1^{(j)}}\sum\limits_{y \in I_2^{(j)}}}_{n\mid x+y}e\left(\frac{l(c\overline{x}-y)}{p}\right)\\ & = &\frac{JHK}{np}+\frac{1}{np}\sum\limits_{l = 1}^{p-1}\sum\limits_{j = 1}^{J}\sum\limits_{x \in I_1^{(j)}}\sum\limits_{y \in I_2^{(j)}} e\left(\frac{l(c\overline{x}-y)}{p}\right)\sum\limits_{s = 1}^{n}e\left(\frac{s(x+y)}{n}\right)+O\left(1\right)\\ & = &\frac{JHK}{np}+\frac{1}{np}\sum\limits_{l = 1}^{p-1}\sum\limits_{s = 1}^{n}\sum\limits_{j = 1}^{J}\sum\limits_{y \in I_2^{(j)}} e\left(\frac{(sp-ln)y}{np}\right)\sum\limits_{x \in I_1^{(j)}}e\left(\frac{sx}{n}+\frac{lc\overline{x}}{p}\right)+O\left(1\right). \end{eqnarray} $ (3.2)

再由柯西不等式以及引理2.3可得

$ \begin{eqnarray} &&\frac{1}{np}\sum\limits_{l = 1}^{p-1}\sum\limits_{s = 1}^{n}\sum\limits_{j = 1}^{J}\sum\limits_{y \in I_2^{(j)}} e\left(\frac{(sp-ln)y}{np}\right)\sum\limits_{x \in I_1^{(j)}}e\left(\frac{sx}{n}+\frac{lc\overline{x}}{p}\right)\\ &\ll&\frac{1}{np}\sum\limits_{l = 1}^{p-1}\sum\limits_{s = 1}^{n}\sum\limits_{j = 1}^{J}\left|\sum\limits_{y \in I_2^{(j)}} e\left(\frac{(sp-ln)y}{np}\right)\right|\cdot\left|\sum\limits_{x \in I_1^{(j)}}e\left(\frac{sx}{n}+\frac{lc\overline{x}}{p}\right)\right|\\ &\ll&\frac{1}{np}\sum\limits_{l = 1}^{p-1}\sum\limits_{s = 1}^{n}\frac{1}{\left<\frac{sp-ln}{np}\right>} \sum\limits_{j = 1}^{J}\left|\sum\limits_{x \in I_1^{(j)}}e\left(\frac{sx}{n}+\frac{lc\overline{x}}{p}\right)\right|\\ &\ll&\frac{1}{np}\sum\limits_{l = 1}^{p-1}\sum\limits_{s = 1}^{n}\frac{1}{\left<\frac{sp-ln}{np}\right>}\cdot J^{\frac{1}{2}}\left( \sum\limits_{j = 1}^{J}\left|\sum\limits_{x \in I_1^{(j)}} e\left(\frac{sx}{n}+\frac{lc\overline{x}}{p}\right)\right|^2\right)^{\frac{1}{2}}\\ &\ll&\frac{1}{np}\sum\limits_{l = 1}^{p-1}\sum\limits_{s = 1}^{n}\frac{1}{\left<\frac{sp-ln}{np}\right>}\cdot J^{\frac{1}{2}}\cdot(p\log^{2}H)^{\frac{1}{2}}, \end{eqnarray} $ (3.3)

其中$ \left<\alpha\right> = \min\left(\{\alpha\}, 1-\{\alpha\}\right) $.记$ m = sp-ln $.则当$ s $取遍模$ n $的完全剩余系, $ l $取遍模$ p $的简化剩余系时, $ m $取遍模$ np $的完全剩余系中与$ p $互素的整数.因此

$ \begin{eqnarray} \sum\limits_{l = 1}^{p-1}\sum\limits_{s = 1}^{n}\frac{1}{\left<\frac{sp-ln}{np}\right>} = \mathop{\sum\limits_{m = 1}^{np}}_{(m, p) = 1} \frac{1}{\left<\frac{m}{np}\right>}\ll np\log(np). \end{eqnarray} $ (3.4)
$ \begin{eqnarray} \end{eqnarray} $ (3.5)

结合(3.1)–(3.4)式, 立即可得

$ \begin{eqnarray*} \sum\limits_{j = 1}^{J}\mathop{\mathop{{\mathop{{\sum}}_{x \in I_1^{(j)}}\mathop{{\sum}}_{y \in I_2^{(j)}}}}_{xy\equiv c (\bmod p)}}_{n\nmid x+y}1 = \left(1-\frac{1}{n}\right)\frac{JHK}{p}+O\left(J^{\frac{1}{2}}p^{\frac{1}{2}}\log p\log H\right). \end{eqnarray*} $

定理1.1证毕.

参考文献
[1] Guy R K. Unsolved problems in number theory[M]. New York: Springer-Verlag, 1981.
[2] Zhang W. On D. H.. Lehmer problem[J]. Chinese Sci. Bull., 1992, 21: 1765–1769.
[3] Zhang W. A problem of D. H.. Lehmer and its generalization (Ⅰ)[J]. Comp. Math., 1993, 86: 307–316.
[4] Zhang W. A problem of D. H.. Lehmer and its generalization (Ⅱ)[J]. Comp. Math., 1994, 91: 47–56.
[5] Zhang W. On the difference between a D. H. Lehmer number and its inverse modulo q[J]. Acta Arith., 1994, 68: 255–263. DOI:10.4064/aa-68-3-255-263
[6] 王晓瑛, 赵秋红. 关于短区间中模q的整数及其逆的分布[J]. 陕西师范大学学报(自然科学版), 2010, 38: 11–14.
[7] Lu Y, Yi Y. On the generalization of the D. H. Lemher problem[J]. Acta Math. Sin. Engl. Ser., 2009, 25: 1269–1274. DOI:10.1007/s10114-009-7652-3
[8] Weil A. Sur les courbes algébriques et les variétés qui s'en déduisent[J]. Actualités Math. Sci., Vol. 1041, Deuxieme Partie, § Ⅳ, Paris, 1945.
[9] Browning T D, Haynes A. Incomplete Kloosterman sums and multiplicative inverses in short intervals[J]. Intern. J. Number The., 2013, 9: 481–486. DOI:10.1142/S1793042112501448