0.1 Additional Topics
Involution, monotonicity
Some interesting problems gives bound for the function for us to use bounding argument.
The key to these kind of problems is to perform certain operations on the function and show that it contradicts with the given criteria.
Here is an interesting problem from the document 169 Functional Equations by Moubinool the showcase the bounding through a specific criteria given by the problem.
Exercise 0.1.1
Find all functions \(f: \mathbb {R} \to \mathbb {R}\) such that \(\left \{ \frac {f(x)}{x} \mid x \in \mathbb {R} \text { and } x \neq 0 \right \}\) is a finite set and \(f(x - 1 - f(x)) = f(x) - x - 1\) for all real \(x\).
Solution.
For this problem, because the terms inside \(f\) in the left-hand side and the terms in the right-hand side are similar, let’s try and make them equal like the following. \[ f(x - 1 - f(x)) = -(x - 1 - f(x)) - 2 \] Let \(\alpha = x - 1 - f(x)\) and we get \(f(\alpha ) = -\alpha - 2\). But there is a problem here. The number of terms for \(\frac {f(x)}{x}\) is bounded, but we have infinitely many \(\frac {-\alpha - 2}{\alpha }\). This tells us that \(\alpha \) is finite for the condition to be satisfied.
Further observing the problem, let’s substitute \(\alpha \) for \(x\). We obtain \begin{align*} f(\alpha - 1 - f(\alpha )) &= -(\alpha - 1 - f(\alpha )) - 2 \\ f(\alpha - 1 + \alpha + 2) &= -\alpha - 1 - \alpha - 2 \end{align*}
and \(f(2\alpha + 1) = -(2\alpha + 1) - 2\). This shows that \(2\alpha + 1\) is also an element of a finite set, say \(A\), that contains \(\alpha \).
Continuing, we can try and substitute \(2\alpha + 1\) for \(x\). We obtain \(f(2\alpha + 1 - 1 + 2\alpha + 1 + 2) = -(2\alpha + 1) - 2 - 2\alpha - 1 - 1\) and \(f(4\alpha + 3) = -(4\alpha + 3) - 2\). This shows that \(4\alpha + 3 \in A\). From the patterns, we could hypothesize that \(2^n \alpha + 2^n - 1 \in A\). To show this, it suffices to show that if \(2^{n-1} \alpha + 2^{n-1} - 1 \in A\), then \(2^n \alpha + 2^n - 1 \in A\). Consider the following equations that we get for substituting \(x = 2^{n-1} \alpha + 2^{n-1} - 1\). \begin{align*} &\quad \ f \left ( 2^{n-1} \alpha + 2^{n-1} - 1 - 1 + \left ( 2^{n-1} \alpha + 2^{n-1} - 1) + 2 \right ) \right ) \\ &= -\left ( 2^{n-1} \alpha + 2^{n-1} - 1 \right ) - 2 - \left ( 2^{n-1} \alpha + 2^{n-1} - 1 \right ) - 1 \\ &\qquad \quad f \left ( 2^n \alpha + 2^n - 1 \right ) = -\left ( 2^n \alpha + 2^n - 1 \right ) - 2 \end{align*}
Indeed by induction, \(2^n \alpha + 2^n - 1 \in A\) for all positive integers \(n\). In other words, \(\alpha = -1\) is the only possible case and \(\alpha = x - 1 - f(x) = -1\). Therefore, \(f(x) = x\) is the only solution.
Some interesting problems don’t give us the options for classic discoveries.
In such problems, it is very important to not underestimate the condition \(f: \mathbb {R} \to \mathbb {R}\).
Here is an example problem where \(f: \mathbb {R} \to \mathbb {R}\) acts as a breakthrough of the solution.
Exercise 0.1.2: 2014 KMO Problem 2 (\( \bullet \bullet \circ \))
Find all functions \(f: \mathbb {R} \to \mathbb {R}\) such that the following equation holds for all \(x, y \in \mathbb {R}\). \[ f(xf(x) + f(x)f(y) + y - 1) = f(xf(x) + xy) + y - 1 \]
Solution.
This problem is very interesting because just by looking at the equation, it is evident that \(f(x) = x\) works. However, it is quite difficult to find anything else other than that, so proving the uniqueness of the solution is the main part of this problem.
First, substitute \(x = 0\) to the equation. Then, we can see that the following equation holds. \[ f(f(0)f(y) + y - 1) = f(0) + y - 1 \] Now how is this supposed to be useful? At glance, it isn’t. There are limited information that we can learn from this equation. However, if you look very closely at the equation, we can see that \(f(0)\) is defined since \(f: \mathbb {R} \to \mathbb {R}\) and \(y - 1\) can be any real number. Meaning, \(f\) is surjective!
From here, we are pretty much done. Because \(f\) is surjective, we can choose some \(a\) such that \(f(a) = 0\). Substituting \(x = a\), we obtain \(f(y - 1) = f(ay) + y - 1\). Therefore, it is natural for us to consider the case when \(ay = y - 1\), which is \(y = \frac {1}{1 - a}\). However, we have to be careful here because we are implying that \(a \neq 1\).
Assume for the contradiction that \(a = 1\) is unique value such that \(f(a) = 0\). Substituting \(x = y = 1\), we obtain \(f(0) = 0\) and \(a = 1\) is not unique if \(f(1) = 0\). In other words, we can choose some \(a \neq 1\) such that \(ay = y - 1\), which will give us \(y - 1 = \frac {a}{1 - a} = 0\). Solving the equation, \(a = 0\) and \(f(0) = 0\).
Substituting \(x = 0\) back to the equation, we get \(f(y - 1) = y - 1\) and \(f(x) = x\) is the unique answer.
Unlike other problems, this one requires some effort to find that \(f(0) = 0\). In this problem \(f: \mathbb {R} \to \mathbb {R}\) was an important information for concluding that \(f\) is surjective.