4
$\begingroup$

It's well known that the continuous solutions to the Cauchy's functional equation: $$ f ( x + y ) = f ( x ) + f ( y ) $$ are of the form $ f ( x ) = c x $ for some constant $ c $. However, I would like to know if the following generalization of the problem is true.

Suppose $ f $ is continuous, and there exist constants $ a < 0 < b $ such that $$ a \le f ( x ) + f ( y ) - f ( x + y ) \le b $$ for every real $ x $ and $y$. Does this condition imply that there is a constant $ c $ such that $$ a \le f ( x ) - c x \le b $$ for every real $ x $?

$\endgroup$
1
  • $\begingroup$ At least case with $f(x)+f(y)-f(x+y)=d$ for $d$ constant should be easy, as it turns into Cauchy's equation by $g(x)=f(x)-d$ and solves to $f(x)=cx+d$. $\endgroup$ – Sil Apr 15 '20 at 12:44
3
$\begingroup$

You can rewrite your inequlities in terms of $ g ( x ) = f ( x ) - \frac { a + b } 2 $ and $ \epsilon = \frac { b - a } 2 $, and ask for the following:

If for some nonnegative real $ \epsilon $ $$ | g ( x + y ) - g ( x ) - g ( y ) | \le \epsilon \tag 0 \label 0 $$ for every real $ x $ and $ y $, is there an additive function $ A $ such that $$ | g ( x ) - A ( x ) | \le \epsilon \tag 1 \label 1 $$ for all $ x $? Is such $ A $ unique? Does continuity of $ g $ imply linearity of $ A $?

The answers to all these questions are positive. In fact this notion is well-known and has a name: stability. A good reference for stability of many famous functional equations is Hyers-Ulam-Rassias stability of functional equations in nonlinear analysis by S.M. Jung. I give the proof on page 21 of that book, with some minor changes.

The trick is to define $ A ( x ) = \lim _ { n \to \infty } 2 ^ { - n } g ( 2 ^ n x ) $. To show that the limit exists for every $ x $, first note that by \eqref{0}, $ | g ( 2 x ) - 2 g ( x ) | \le \epsilon $, or equivalently $ \big| \frac 1 2 g ( x ) - g \big( \frac x 2 \big) \big| \le \frac \epsilon 2 $ for every $ x $. It follows that $$ \left| 2 ^ { - n } g ( x ) - g \left( 2 ^ { - n } x \right) \right| = \left| \sum _ { i = 0 } ^ { n - 1 } \Big( 2 ^ { - n + i } g \big( 2 ^ { - i } x \big) - 2 ^ { - n + i + 1 } g \big( 2 ^ { - i - 1 } x \big) \Big) \right| \\ \le \sum _ { i = 0 } ^ { n - 1 } 2 ^ { - n + i + 1 } \big| 2 ^ { - 1 } g \big( 2 ^ { - i } x \big) - g \big( 2 ^ { - i - 1 } x \big) \big| \le \sum _ { i = 0 } ^ { n - 1 } 2 ^ { - n + i } \epsilon = ( 1 - 2 ^ { - n } ) \epsilon \text . \tag 2 \label 2 $$ Thus for $ m < n $ we get $$ | 2 ^ { - m } g ( 2 ^ m x ) - 2 ^ { - n } g ( 2 ^ n x ) | = 2 ^ { - m } | g ( 2 ^ m x ) - 2 ^ { m - n } g ( 2 ^ n x ) | \le 2 ^ { - m } ( 1 - 2 ^ { m - n } ) \epsilon = ( 2 ^ { - m } - 2 ^ { - n } ) \epsilon $$ which shows that $ \big( 2 ^ { - n } g ( 2 ^ n x ) \big) _ { n = 0 } ^ \infty $ is a Cauchy sequence and hence convergent. It follows from \eqref{0} that $ | g ( 2 ^ n x + 2 ^ n y ) - g ( 2 ^ n x ) - g ( 2 ^ n y ) | \le \epsilon $. Dividing by $ 2 ^ n $ and letting $ n \to \infty $ we see that $ A $ is an additive function. If we replace $ x $ by $ 2 ^ n x $ in \eqref{2} and take the limit, we have the inequality \eqref{1}.

Suppose that $ B $ is another additive function satisfying \eqref{1}. We can see that $$ | A ( x ) - B ( x ) | = \frac 1 n | A ( n x ) - B ( n x ) | \le \frac 1 n | A ( n x ) - g ( n x ) | + \frac 1 n | g ( n x ) - B ( n x ) | \le \frac { 2 \epsilon } n \text . $$ Hence $ B = A $, and $ A $ is the unique additive function satisfying the inequality \eqref{1}.

At last, we show that if $ g $ is continuous at any point $ x $, then $ A $ is continuous at $ 0 $, and since it's additive, continuous everywhere, which shows that it is linear. Since $ g $ is continuous at $ x $, there is a positive $ \delta $ such that if $ | y | < \delta $ then $ | g ( x + y ) - g ( x ) | < \epsilon $. We then have by \eqref{1} $$ | A ( y ) | = | A ( x + y ) - A ( x ) | \\ \le | A ( x + y ) - g ( x + y ) | + | g ( x + y ) - g ( x ) | + | g ( x ) - A ( x ) | < 3 \epsilon \text . $$ Since $ A $ is additive, we get that if $ | y | < \frac \delta n $ then $ | A ( y ) | < \frac { 3 \epsilon } n $, which shows that $ A $ is continuous at $ 0 $, and we're done.

Jung, Soon-Mo, Hyers-Ulam-Rassias stability of functional equations in nonlinear analysis, Springer Optimization and Its Applications 48. Berlin: Springer (ISBN 978-1-4419-9636-7/hbk; 978-1-4419-9637-4/ebook). xiii, 362 p. (2011). ZBL1221.39038.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.