3
$\begingroup$

A Baire 1 function $f:[0,1]\rightarrow \mathbb{R}$ need not be bounded. However, thanks to the Baire category theorem, we know there is $N\in \mathbb{N}$ and a sub-interval $(c, d) \subset [0,1]$ such that $(\forall x\in (c, d))( |f(x)|\leq N)$. In other words, a Baire 1 function may not be bounded on the unit interval, we know it is bounded on some sub-interval.

Similarly, a Baire 1 function $f:[0,1]\rightarrow \mathbb{R}^+$ may have infimum equal to $0$, but there is always some sub-interval $(c, d)\subset [0,1]$ where $f$ has non-zero infimum.

I am curious what other nice properties Baire 1 functions can have that fail on the unit interval, but do hold on some sub-interval. I would also be happy with examples for e.g. upper semi-continuous functions.

$\endgroup$
1
  • 1
    $\begingroup$ Probably not relevant towards an answer, but a possibly interesting parallel: (1) If $f$ is Baire $1,$ then $f$ is bounded on some subset that is an interval with positive length. (2) If $f$ is Lebesgue measurable, then $f$ is bounded on some subset that is closed with positive Lebesgue measure (by Luzin's theorem). As an after-thought, in (2) "positive Lebesgue measure" can be improved to "positive Borel measure", and even to "positive Jordan content", although this doesn't mathematically strengthen the result. Also, "closed set" can be improved to "nowhere dense perfect set" . . . $\endgroup$ Commented Feb 21, 2023 at 18:33

1 Answer 1

5
$\begingroup$

Let $ s:[0;1]\to[0;1]$ be the average binary digit partial function defined as follows:

$$ s(x)\ :=\ \lim_{n=\infty}\quad s_n(x) $$

where $$ s_n(x)\, :=\,\ \frac 1n\cdot \sum_{k=1}^n b_{-k}(x) $$ and $\ b_k(x)\ $ are the binary digits of $\ x.$

Borel's theorem tells us that $\ s(x)=\frac12\ $ for all $\ x\in[0;1]\ $ except for Lebesgue measure $\ 0.$ Nevertheless, the following popular question was open for a rather long time:

$\qquad$ does there exist $\ x\in(0;1)\ $ (an open interval)

$\qquad\qquad\qquad$ such that $\ s(x)=x\,?$

I answered this question and more in the year 1959; in particular, I replaced function $\ f(x):=x\ $ by any function from a larger set of functions:

Theorem   Let $\ f:[0;1]\to[0;1]\ $ be an arbitrary Baire 1 function. Then for arbitrary $ \emptyset \ne [a;b]\subseteq [0;1] $ there exist continuum many $\ x\in[a;b]\ $ such that $$ s(x)\ =\ f(x). $$

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .