2

Find all functions $ f : \mathbb R \to \mathbb R $ such that for all $ a , b \in \mathbb R$: $$ f ( a ) + f \big( a + f ( b ) \big) = b + f \big( f ( a ) + f ^ 2 ( b ) \big) \text . $$

Here, for any $ n \in \mathbb N $, $ f ^ n $ denotes the $ n $-th iteration of $ f $.


My ideas so far:

I substituted $ ( 0 , x ) $ that yields: $$ f ( 0 ) + f ^ 2 ( x ) = x + f \big( f ( 0 ) + f ^ 2 ( x ) \big) \text . \tag 1 \label 1 $$

Let's say that $ a , b \in \mathbb R $, $ a \ne b $ and $ f ( a ) = f ( b ) $. Then the LHS does not change value with $ a , b $ but the RHS does. That is a contradiction and thus $ f ( a ) = f ( b ) \implies a = b $. The function is therefore injective.

If we substitute $ \big( x , f ( x ) \big) $ we can cross the functions like so: $$ f \big( x + f ^ 2 ( x ) \big) = f \big( f ( x ) + f ^ 3 ( x ) \big) \text ; $$ $$ \therefore \quad x + f ^ 2 ( x ) = f ( x ) + f ^ 3 ( x ) \text . \tag 2 \label 2 $$

If we assume $ f ( 0 ) = 0 $, we have $ f ^ 2 ( x ) = x + f ^ 3 ( x ) $.
Using \eqref{2} we get $ f ( x ) = 2 x $. However, this does not satisfy the functional equation and thus we can conclude that $ f ( 0 ) \ne 0 $.

I also noticed that if you substitute $ f ( x ) $ for $ x $, you get $ f ( x ) + f ^ 3 ( x ) =f ^ 2 ( x ) + f ^ 4 ( x ) $ and expressing $ f ^ 3 ( x ) $ from \eqref{2} we get an intresting result: $ f ^ 4 ( x ) = x $ which means the function is iterative with a cycle of $ 4 $ (or $ 2 $ or $ 1 $).

I am not sure how to continue or what substitution should I try next.

  • Can you elaborate on "LHS does not change value with a, b"? – Calvin Lin Nov 30 '20 at 22:20
  • You have exactly one x "outside" f, which means that if a=0 and b remains distinct as a variable, the LHS of (1) remains constant with changing b=x, while the RHS changes since the variable is outside the function. Therefore a and b cannot be distinct, i.e. it cannot happen that one is constant and the other free. The "one x outside the function means injective" is a pretty common shortcut in functional equations as far as I know. – mathcactus Nov 30 '20 at 22:47
  • Thanks for clarifying. That's what I thought, but it was slightly confusing with the variables, so I wanted to ensure that's what you intended. It's often much better to just write it out as $ a+ f(f(0 + f^2 (a) ) = f(0) + f^2 (a) = f(0) + f^2(b) = \ldots $. – Calvin Lin Dec 01 '20 at 00:42
  • Can you elaborate on what you're doing after establishing equation (2)? IE How does setting $f(0) = 0 $ result in $f^2(x) = x + f^3(x)$? -> Ah nevermind, you were substituting this into (1). FYI It seems like you have the work, but have shortened it in such a manner that we need to guess at what you've done, which isn't good. – Calvin Lin Dec 01 '20 at 00:46
  • https://artofproblemsolving.com/community/c6h1574379p9683609 – zwim Dec 01 '20 at 01:12

1 Answers1

0

You can show that there is no function $ f : \mathbb A \to \mathbb A $ satisfying $$ f ( x ) + f \big( x + f ( y ) \big) = y + f \Big( f ( x ) + f \big( f ( y ) \big) \Big) \tag 0 \label 0 $$ for all $ x , y \in \mathbb A $, where $ ( \mathbb A , + ) $ is any abelian group with the neutral element $ 0 $ and the inverse function $ - $, such that there is $ b \in \mathbb A $ with $ 5 b \ne 0 $. As you are interested in $ \mathbb A = \mathbb R $ with the group operation $ + $ taken to be the usual addition of real numbers, this would be the case, since any nonzero real number can be chosen as $ b $.

To see this, substitute $ f ( x ) $ for $ x $, and see that $$ f \big( f ( x ) \big) - y = f \Big( f \big( f ( x ) \big) + f \big( f ( y ) \big) \Big) - f \big( f ( x ) + f ( y ) \big) \\ = f \Big( f \big( f ( y ) \big) + f \big( f ( x ) \big) \Big) - f \big( f ( y ) + f ( x ) \big) = f \big( f ( y ) \big) - x \text , $$ which in particular shows $$ f \big( f ( x ) \big) = f \big( f ( 0 ) \big) - x \text . \tag 1 \label{1a} $$ Putting $ x = 0 $ in \eqref{0} and using \eqref{1a} you have $$ f ( 0 ) + f \big( f ( 0 ) \big) - y = y + f \Big( f ( 0 ) + f \big( f ( 0 ) \big) - y \Big) \text , $$ which by letting $ a = f ( 0 ) + f \big( f ( 0 ) \big) $ and substituting $ - x + a $ for $ y $ shows that $$ f ( x ) = 2 x - a \text . \tag 2 \label{2a} $$ Using \eqref{1a} and \eqref{2a} you get $ 5 x = 0 $ for all $ x \in \mathbb A $, and in particular for $ x = b $, which is a contradiction.


In case every $ b \in \mathbb A $ is of order $ 5 $, choosing any $ a \in \mathbb A $ and taking $ f $ to be of the form \eqref{2a}, the equation \eqref{0} will be satisfied for all $ x , y \in \mathbb A $. To see this, use \eqref{2a} to rewrite \eqref{0} as $$ 2 x - a + 2 ( x + 2 y - a ) - a = y + 2 \big( 2 x - a + 2 ( 2 y - a ) - a \big) - a \text , $$ or equivalently $$ 4 x + 4 y - 4 a = 4 x + 9 y - 9 a \text , $$ which is true since $ 5 y = 0 $ and $ 5 a = 0 $. As the derivations resulting in \eqref{2a} were valid for any abelian group (independent of the order of its elemnts), we've characterized all the solutions it this case. Examples of abelian groups in which the order of every element is $ 5 $ are trivial groups, the cyclic group $ \mathbb Z _ 5 $ and the direct product of $ \mathbb Z _ 5 $ with itself.