Maths Olympiad Prep

Library / /520 of 520

Geometry Difficulty 8.1 Shortlist Prove it

Example 37([31.6]) Prove: There exists a convex 1990-gon that simultaneously has the following properties:
(i) All interior angles are equal;
(ii) The lengths of the 1990 sides are a permutation of the numbers 12,22,32,,19892,199021^{2}, 2^{2}, 3^{2}, \cdots, 1989^{2}, 1990^{2}.

Solution

Suppose this polygon exists, with vertices A1,A2,,A1990A_{1}, A_{2}, \cdots, A_{1990}. Choose a Cartesian coordinate system, placing vertex A1990A_{1990} at the origin and edge A1990A1A_{1990} A_{1} along the xx-axis. From condition (i), we know that each interior angle of this convex polygon is 1θ1-\theta, where θ=2π/1990\theta=2 \pi / 1990 (why). Considering the edges A1A2,A2A3,,A1989A1990,A1990A1A_{1} A_{2}, A_{2} A_{3}, \cdots, A_{1989} A_{1990}, A_{1990} A_{1} as vectors, we have (why)
A1A2+A2A3++AjAj+1++A1989A1990+A1990A1=0.\overrightarrow{A_{1} A_{2}}+\overrightarrow{A_{2} A_{3}}+\cdots+\overrightarrow{A_{j} A_{j+1}}+\cdots+\overrightarrow{A_{1989} A_{1990}}+\overrightarrow{A_{1990} A_{1}}=\overrightarrow{0} .
Expressing these vectors in complex form, we get (why)
AjAj+1=rj2exp(ijθ),1j1990,\overrightarrow{A_{j} A_{j+1}}=r_{j}^{2} \exp (\mathrm{i} j \theta), \quad 1 \leqslant j \leqslant 1990,

where i=1\mathrm{i}=\sqrt{-1}, rj2r_{j}^{2} is the length of edge AjAj+1A_{j} A_{j+1}, and A1991=A1A_{1991}=A_{1}. From equations (1) and (2), we obtain
T=1j1990rj2exp(ijθ)=0T=\sum_{1 \leqslant j \leqslant 1990} r_{j}^{2} \exp (\mathrm{i} j \theta)=0

From condition (2), we know that r1,r2,,r1990r_{1}, r_{2}, \cdots, r_{1990} should be a permutation of 1,2,,19901,2, \cdots, 1990. Thus, our problem is whether and how to select a permutation of 1,2,,19901,2, \cdots, 1990 such that equation (3) holds.

The sum in equation (3) is typically called an exponential sum. Therefore, we need knowledge about when an exponential sum equals zero. The only known result in this regard is: for any integer M>1M > 1,
1kMexp(2πik/M)=0\sum_{1 \leqslant k \leqslant M} \exp (2 \pi \mathrm{i} k / M)=0

From this, we can deduce that for any integer cc coprime with MM, we have (why)
1kMexp(2πick/M)=0\sum_{1 \leqslant k \leqslant M} \exp (2 \pi i c k / M)=0

To use these conclusions to solve our problem, we need to select r1,r2,,r1990r_{1}, r_{2}, \cdots, r_{1990} in a special way so that the sum in equation (3) can be expressed as a combination of the sums in equations (4) and (5). This requires considering how to appropriately combine the terms in the sum in equation (3) corresponding to the angles jθj \theta to form the sums in equations (4) and (5).

Since the order of the angles jθj \theta is fixed, this complicates the combination of terms. Given that for any integer dd,
exp(ijθ)=exp(i(j+1990d)θ)\exp (\mathrm{i} j \theta)=\exp (\mathrm{i}(j+1990 d) \theta)

we can find a complete residue system modulo 1990, u1,u2,,u1990u_{1}, u_{2}, \cdots, u_{1990}, such that (why)
T=1j1990rj2exp(iθuj)=0T=\sum_{1 \leq j \leqslant 1990} r_{j}^{2} \exp \left(\mathrm{i} \theta u_{j}\right)=0

Thus, the problem becomes how to choose r1,r2,,r1990r_{1}, r_{2}, \cdots, r_{1990} and u1,u2,,u1990u_{1}, u_{2}, \cdots, u_{1990} so that by appropriately combining and summing them, we can achieve our goal. This involves using the properties of residue systems (see Chapter 3, §2). Guided by our approach, noting the property eix=1\mathrm{e}^{\mathrm{i} x}=-1 and the characteristics of the number 1990, we select {rj}\left\{r_{j}\right\} in the form ak,m=10k+m(1m10,0k198)a_{k, m}=10 k+m(1 \leqslant m \leqslant 10,0 \leqslant k \leqslant 198) and {uj}\left\{u_{j}\right\} in the form bk,n=10k+199n(0n9,0k198)b_{k, n}=10 k+199 n(0 \leqslant n \leqslant 9,0 \leqslant k \leqslant 198). In summing TT, they are combined as follows:
T=0k198TkT=\sum_{0 \leqslant k \leqslant 198} T_{k}

where
Tk=ak,12exp(iθbk,0)+ak,62exp(iθbk,5)+ak,22exp(iθbk,2)+ak,72exp(iθbk,7)+ak,32exp(iθbk,4)+ak,82exp(iθbk,9)+ak,42exp(iθbk,6)+ak,92exp(iθbk,1)+ak,52exp(iθbk,8)+ak,102exp(iθbk,3)\begin{aligned} T_{k}= & a_{k, 1}^{2} \exp \left(\mathrm{i} \theta b_{k, 0}\right)+a_{k, 6}^{2} \exp \left(\mathrm{i} \theta b_{k, 5}\right)+a_{k, 2}^{2} \exp \left(\mathrm{i} \theta b_{k, 2}\right) \\ & +a_{k, 7}^{2} \exp \left(\mathrm{i} \theta b_{k, 7}\right)+a_{k, 3}^{2} \exp \left(\mathrm{i} \theta b_{k, 4}\right)+a_{k, 8}^{2} \exp \left(\mathrm{i} \theta b_{k, 9}\right) \\ & +a_{k, 4}^{2} \exp \left(\mathrm{i} \theta b_{k, 6}\right)+a_{k, 9}^{2} \exp \left(\mathrm{i} \theta b_{k, 1}\right) \\ & +a_{k, 5}^{2} \exp \left(\mathrm{i} \theta b_{k, 8}\right)+a_{k, 10}^{2} \exp \left(\mathrm{i} \theta b_{k, 3}\right) \end{aligned}

It is not difficult to see that
Tk=exp(10kθi){(ak,12ak,62)exp(02πi/5)+(ak,22ak,72)exp(12πi/5)+(ak,32ak,82)exp(22πi/5)+(ak,42ak,92)exp(32πi/5)+(ak,52ak,102)exp(42πi/5)}=5exp(10kθi){(20k+7)exp(02πi/5)+(20k+9)exp(12πi/5)+(20k+11)exp(22πi/5)+(20k+13)exp(32πi/5)+(20k+15)exp(42πi/5)}=5exp(10kθi){1exp(02πi/5)+3exp(12πi/5)+5exp(22πi/5)+7exp(32πi/5)+9exp(42πi/5)}=5Aexp(10kθi)=5Aexp(2kπi/199)\begin{aligned} T_{k}= & \exp (10 k \theta \mathrm{i})\left\{\left(a_{k, 1}^{2}-a_{k, 6}^{2}\right) \exp (0 \cdot 2 \pi \mathrm{i} / 5)+\left(a_{k, 2}^{2}-a_{k, 7}^{2}\right) \exp (1 \cdot 2 \pi \mathrm{i} / 5)\right. \\ & +\left(a_{k, 3}^{2}-a_{k, 8}^{2}\right) \exp (2 \cdot 2 \pi \mathrm{i} / 5)+\left(a_{k, 4}^{2}-a_{k, 9}^{2}\right) \exp (3 \cdot 2 \pi \mathrm{i} / 5) \\ & \left.+\left(a_{k, 5}^{2}-a_{k, 10}^{2}\right) \exp (4 \cdot 2 \pi \mathrm{i} / 5)\right\} \\ = & -5 \exp (10 k \theta \mathrm{i})\{(20 k+7) \exp (0 \cdot 2 \pi \mathrm{i} / 5)+(20 k+9) \exp (1 \cdot 2 \pi \mathrm{i} / 5) \\ & +(20 k+11) \exp (2 \cdot 2 \pi \mathrm{i} / 5)+(20 k+13) \exp (3 \cdot 2 \pi \mathrm{i} / 5) \\ & +(20 k+15) \exp (4 \cdot 2 \pi \mathrm{i} / 5)\} \\ = & -5 \exp (10 k \theta \mathrm{i})\{1 \exp (0 \cdot 2 \pi \mathrm{i} / 5)+3 \exp (1 \cdot 2 \pi \mathrm{i} / 5) \\ & +5 \exp (2 \cdot 2 \pi \mathrm{i} / 5)+7 \exp (3 \cdot 2 \pi \mathrm{i} / 5)+9 \exp (4 \cdot 2 \pi \mathrm{i} / 5)\} \\ = & -5 A \exp (10 k \theta \mathrm{i})=-5 A \exp (2 k \pi \mathrm{i} / 199) \end{aligned}

Therefore,
T=5A0<k198exp(2kπi/199)=0T=-5 A \sum_{0<k \leq 198} \exp (2 k \pi \mathrm{i} / 199)=0

This specifically constructs the required polygon.

Want a route through all this instead of an archive? The track puts 2,000 problems in a working order, from AMC 10 level to the IMO shortlist.

Source: NuminaMath-1.5, licensed Apache-2.0. Statement and solution reproduced as published; topic and difficulty added by this site.