Maths Olympiad Prep

Library / /7 of 10

, 2023

Algebra Difficulty 8.8 Shortlist Prove it China

(1) Prove: on the complex plane, for each line passing through the origin except for the real axis, there is at most one point zz such that 1+z23z64\frac{1+z^{23}}{z^{64}} is a real number.

(2) Prove: for any complex number a0a \ne 0 and any real number θ\theta, the equation
1+z23+az64=0 1 + z^{23} + a z^{64} = 0
has at least one root in the set
Sθ={zCRe(zeiθ)zcosπ20}. S_{\theta} = \{z \in \mathbb{C} \mid \mathrm{Re}(z \cdot e^{-i\theta}) \ge |z| \cdot \cos \frac{\pi}{20}\}.

Solution

(1) For a non-zero complex number zz, let z=r(cosθ+isinθ)z = r(\cos\theta + i\sin\theta) where r>0r > 0 and θ[0,2π)\theta \in [0, 2\pi), we have
f(z):=1+z23z64=1r64(cos(64θ)+isin(64θ)+r23(cos(41θ)+isin(41θ))). f(z) := \frac{1+z^{23}}{z^{64}} = \frac{1}{r^{64}} (\cos(-64\theta) + i\sin(-64\theta) + r^{23}(\cos(-41\theta) + i\sin(-41\theta))).
Therefore,
1+z23z64RIm1+z23z64=164(sin64θ+r23sin41θ)=0sin64θ+r23sin41θ=0. \begin{align*} \frac{1+z^{23}}{z^{64}} \in \mathbb{R} &\Leftrightarrow \mathrm{Im} \frac{1+z^{23}}{z^{64}} = \frac{-1}{64} (\sin 64\theta + r^{23} \sin 41\theta) = 0 \\ &\Leftrightarrow \sin 64\theta + r^{23} \sin 41\theta = 0. \end{align*}
For θ0,π\theta \neq 0, \pi, when sin41θ=0\sin 41\theta = 0, we have sin64θ0\sin 64\theta \neq 0 and there are no solutions to the equation.
When sin41θ0\sin 41\theta \neq 0, we have
r=g0(θ):=(sin64θsin41θ)1/23, r = g_0(\theta) := \left( \frac{-\sin 64\theta}{\sin 41\theta} \right)^{1/23},
and for r>0r > 0, sin64θ\sin 64\theta and sin41θ\sin 41\theta must have different signs. Noting that g0(θ)=g0(θ+π)g_0(\theta) = -g_0(\theta + \pi), we can conclude that the proposition holds.

For a fixed λ[0,π)\lambda \in [0, \pi), we consider the set of points on the complex plane
Γλ:={zf(z)eiλR}={zsin(λ+64θ)+r23sin(λ+41θ)=0}. \Gamma_{\lambda} := \{z \mid f(z)e^{-i\lambda} \in \mathbb{R}\} = \{z \mid \sin(\lambda + 64\theta) + r^{23}\sin(\lambda + 41\theta) = 0\}.
The problem can be divided into two cases:
(1°) λ+64θλ+41θ0(modπ)\lambda + 64\theta \equiv \lambda + 41\theta \equiv 0 \pmod{\pi}, then θkπ23(modπ)\theta \equiv \frac{k\pi}{23} \pmod{\pi} and λ5kπ23(modπ)\lambda \equiv \frac{5k\pi}{23} \pmod{\pi} hold for some integer kk.
(2°) If the above equation does not hold, then similar to the discussion in the first question: There is at most one point zz on every line passing through the origin that lies in Γλ\Gamma_{\lambda}. (Specifically, there is at most one intersection point between any ray originating from the origin and Γλ\Gamma_{\lambda})

gλ(θ):=(sin(λ+64θ)sin(λ+41θ))1/23. g_{\lambda}(\theta) := \left( \frac{-\sin(\lambda + 64\theta)}{\sin(\lambda + 41\theta)} \right)^{1/23}.
As previously mentioned, we want to prove that when zz is in the sector corresponding to some interval of θ\theta, the set composed of all points in this interval that make gλ(θ)>0g_{\lambda}(\theta) > 0,
()eiλf(z)=1(gλ(θ))64(cos(λ+64θ)+r23cos(λ+41θ)) takes all real values. (**) \quad -e^{-i\lambda} f(z) = - \frac{1}{(g_{\lambda}(\theta))^{64}} (\cos(\lambda + 64\theta) + r^{23} \cos(\lambda + 41\theta)) \text{ takes all real values.}
The θ\theta that satisfies gλ(θ)>0g_{\lambda}(\theta) > 0 is made up of several subintervals. At the endpoints of these intervals, there are two possible cases:
a) The θ\theta at the endpoint satisfies λ+41θ0(modπ)\lambda + 41\theta \equiv 0 \pmod{\pi}, but λ+64θ≢0(modπ)\lambda + 64\theta \not\equiv 0 \pmod{\pi}. In this case, as θ\theta approaches this endpoint within the interval, gλ(θ)g_{\lambda}(\theta) \to \infty and the corresponding a0a \to 0 (ρ0\rho \to 0);
b) The θ\theta at the endpoint satisfies λ+64θ0(modπ)\lambda + 64\theta \equiv 0 \pmod{\pi}, but λ+41θ≢0(modπ)\lambda + 41\theta \not\equiv 0 \pmod{\pi}. In this case, as θ\theta approaches this endpoint within the interval, gλ(θ)0g_{\lambda}(\theta) \to 0 and the corresponding aa \to \infty (when cos(λ+64θ)\cos(\lambda + 64\theta) at the endpoint is 1-1, ρ+\rho \to +\infty; when it's 1, ρ\rho \to -\infty).
Now we proceed to prove that for any interval of length π10\frac{\pi}{10} (in terms of θ\theta), there exists a subinterval in which (***) holds.
1) If there exists θ0\theta_0 in the interval (with an appropriate λ\lambda) satisfying the aforementioned condition 1°, then the distance from one of the endpoints of the interval (without loss of generality, assuming it to be the right endpoint) to θ0\theta_0 is not less than π20>2π41>3π64\frac{\pi}{20} > \frac{2\pi}{41} > \frac{3\pi}{64}. Therefore, during the process of changing from θ0\theta_0 to this endpoint, the following two situations may occur (also without loss of generality, assuming that sin(λ+64θ)\sin(\lambda + 64\theta) takes positive values in (θ0,θ0+π64)(\theta_0, \theta_0 + \frac{\pi}{64})). One case is that

| | θ0\theta_0 | | θ0+π64\theta_0 + \frac{\pi}{64} | | θ0+π41\theta_0 + \frac{\pi}{41} | | θ0+π32\theta_0 + \frac{\pi}{32} | | θ0+3π64\theta_0 + \frac{3\pi}{64} | | θ0+2π41\theta_0 + \frac{2\pi}{41} |
|----------------|------------|---|------------------------------|---|-------------------------------|---|-------------------------------|---|-------------------------------|---|-------------------------------|
| sin(λ+64θ)\sin(\lambda + 64\theta) | 0 | + | 0 | - | - | - | 0 | + | 0 | - | - |
| sin(λ+41θ)\sin(\lambda + 41\theta) | 0 | - | - | - | 0 | + | + | + | + | + | 0 |
| cos(λ+64θ)\cos(\lambda + 64\theta) | ±1\pm1 | | 1\mp1 | | | | ±1\pm1 | | 1\mp1 | | |

At this point, we observe that the piecewise continuous function ρ\rho (in terms of θ\theta) satisfies the following conditions:
limθ(θ0+π32)ρ=±,limθ(θ0+3π64)+ρ=,ρ(θ0+π41)=ρ(θ0+2π41)=0. \lim_{\theta \to (\theta_0 + \frac{\pi}{32})^{-}} \rho = \pm\infty, \quad \lim_{\theta \to (\theta_0 + \frac{3\pi}{64})^{+}} \rho = \mp\infty, \quad \rho(\theta_0 + \frac{\pi}{41}) = \rho(\theta_0 + \frac{2\pi}{41}) = 0.
Therefore, on the interval [θ0+π41,θ0+3π32)(θ0+3π64,θ0+2π41][\theta_0 + \frac{\pi}{41}, \theta_0 + \frac{3\pi}{32}) \cup (\theta_0 + \frac{3\pi}{64}, \theta_0 + \frac{2\pi}{41}], ρ\rho can take on any real value.
The other case is that

| | θ0\theta_0 | | θ0+π64\theta_0 + \frac{\pi}{64} | | θ0+π41\theta_0 + \frac{\pi}{41} | | θ0+π32\theta_0 + \frac{\pi}{32} | | θ0+3π64\theta_0 + \frac{3\pi}{64} | | θ0+2π41\theta_0 + \frac{2\pi}{41} |
|----------------|------------|---|------------------------------|---|-------------------------------|---|-------------------------------|---|-------------------------------|---|-------------------------------|
| sin(λ+64θ)\sin(\lambda + 64\theta) | 0 | + | 0 | - | - | - | 0 | + | 0 | - | - |
| sin(λ+41θ)\sin(\lambda + 41\theta) | 0 | + | + | + | 0 | - | - | - | - | - | 0 |
| cos(λ+64θ)\cos(\lambda + 64\theta) | ±1\pm1 | | 1\mp1 | | | | ±1\pm1 | | 1\mp1 | | |

At this point, we observe that ρ\rho can take on any real value on the interval (θ0+π32,θ0+3π64)(\theta_0 + \frac{\pi}{32}, \theta_0 + \frac{3\pi}{64}).
2) For the aforementioned case 2°, considering the values of θ\theta (and λ\lambda) within this interval, we observe that it contains at least 4 zeros of sin(λ+41θ)\sin(\lambda + 41\theta). Furthermore, noting that 343>464\frac{3}{43} > \frac{4}{64}, it follows that among these four zeros of sin(λ+41θ)\sin(\lambda + 41\theta), there must exist two adjacent zeros, denoted as θ1\theta_1 and θ1+π41\theta_1 + \frac{\pi}{41}, between which there are two zeros of sin(λ+64θ)\sin(\lambda + 64\theta), namely θ2\theta_2 and θ2+π64\theta_2 + \frac{\pi}{64}. Without loss of generality, let's assume that θ1+2π41\theta_1 + \frac{2\pi}{41} is also within the interval. In this case, either we have (assuming, without loss of generality, that sin(λ+64θ)\sin(\lambda + 64\theta) takes positive values in (θ1,θ2)(\theta_1, \theta_2)).

| | θ1\theta_1 | θ2\theta_2 | θ2+π64\theta_2 + \frac{\pi}{64} | θ1+π41\theta_1 + \frac{\pi}{41} | θ2+π32\theta_2 + \frac{\pi}{32} |
|----------------|------------|------------|------------------------------|------------------------------|------------------------------|
| sin(λ+64θ)\sin(\lambda + 64\theta) | + | + | 0 | 0 | + |
| sin(λ+41θ)\sin(\lambda + 41\theta) | 0 | - | - | - | 0 |
| cos(λ+64θ)\cos(\lambda + 64\theta) | | ±1\pm1 | 1\mp1 | | ±1\pm1 |

in which case one takes [θ1,θ2)(θ2+π64,θ1+π41][\theta_1, \theta_2) \cup (\theta_2 + \frac{\pi}{64}, \theta_1 + \frac{\pi}{41}] as the desired interval; or we have

| | θ1\theta_1 | θ2\theta_2 | θ2+π64\theta_2 + \frac{\pi}{64} | θ1+π41\theta_1 + \frac{\pi}{41} | θ2+π32\theta_2 + \frac{\pi}{32} |
|----------------|------------|------------|------------------------------|------------------------------|------------------------------|
| sin(λ+64θ)\sin(\lambda + 64\theta) | + | + | 0 | 0 | + |
| sin(λ+41θ)\sin(\lambda + 41\theta) | 0 | + | + | 0 | - |
| cos(λ+64θ)\cos(\lambda + 64\theta) | | ±1\pm1 | 1\mp1 | | ±1\pm1 |

in which case one takes (θ2,θ2+π64)(\theta_2, \theta_2 + \frac{\pi}{64}) as the desired interval.
In conclusion, the proposition is proven. □

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: MathNet, licensed CC-BY-4.0. Statement and solution reproduced as published; topic and difficulty added by this site.