Let $a<b.$ We say that a function $f:[a,b]\to\mathbb{R}$ is Baire $1$ if it is a pointwise limit of some sequence of real-valued continuous functions.

Given a sequence $(g_n)_{n=1}^\infty$ on $[a,b],$ we say that $(g_n)_{n=1}^\infty$ converges to $f$ uniformly if for every $\varepsilon>0,$ there exists $N\in\mathbb{N}$ such that for any $n\geq N$ and any $x\in [a,b],$ we have $$|g_n(x)-f(x)|<\varepsilon.$$

Theorem: If $(g_n)_{n=1}^\infty$ is a sequence of real-valued Baire $1$ functions on $[a,b]$ that converges to $f$ uniformly, then $f$ is also Baire $1.$

My attempt: For each $n\in\mathbb{N},$ since $g_n$ is Baire $1,$ there exists a sequence of continuous functions $(g_n^m)_{m=1}^\infty$ that converges to $g_n$ pointwise. Let $$f_n=g_n^n,$$ that is, choose $f_n$ 'diagonally' from the pool of $g_n^m.$ However, I notice that such $(f_n)_{n=1}^\infty$ may not converge to $f$ pointwise, as for any $x\in [a,b]$ and $\varepsilon>0,$ by uniform convergence, there exists $N\in\mathbb{N}$ such that $$|g_N(x)-f(x)|<\frac{\varepsilon}{2}.$$ By pointwise limit of $g_N,$ there exists $M \in \mathbb{N}$ such that for any $n\geq M,$ we have $$|g_N(x) - g_{N}^m(x)|<\frac{\varepsilon}{2}.$$ However, it may be the case that $N< M,$ that is, $f_N=g_N^N$ may not be 'captured' by pointwise convergence.

But I cannot change my $f_N$ as this will make the argument circular.

Any hint to fix this problem? If someone wants to provide another shorter proof, I would be glad to have a look.


You need to work with a subsequence that converges at a controllable, fast rate.

The sequence $(g_n)$ converges uniformly to $f$ and has a subsequence $(h_n)$ such that $|h_n(x) - f(x)| < 2^{-(n+1)}$for all $n$ and $x \in [a,b]$.

Using a telescoping sum we have for every $x$,

$$\tag{*}f(x) = h_1(x) + \sum_{n=1}^\infty [h_{n+1}(x) - h_{n}(x)].$$

Note that $h_{n+1} - h_{n}$ as a difference of Baire 1 functions is also Baire 1. Furthermore we have that $|h_{n+1}(x) - h_n(x)| \leqslant 2^{-n}.$

If we can show that the sum on the RHS of (*) is Baire 1 we are done. This can be accomplished by producing a convergent sequence of continuous functions.

Since $h_{n+1} - h_n$ is Baire 1 and bounded, there is a sequence of continuous functions $\phi_{nk}$ such that $\lim_{k \to \infty} \phi_{nk}(x) = h_{n+1}(x) - h_n(x)$ pointwise, and $|\phi_{nk}(x)| \leqslant 2^{-n}$ for all $k$ and $x \in [a,b]$ (details omitted).

For each $k$, there exists a function $\psi_k$ such that

$$\psi_k(x) = \sum_{n=1}^\infty \phi_{nk}(x),$$

where the series converges uniformly for all $x \in [a,b]$. Hence, the functions $\psi_k$ are continuous.

Note that for any positive integer $N$

$$\tag{**} \left|\psi_k(x) - \sum_{n=1}^\infty[h_{n+1}(x) - h_n(x)]\right| \\ \leqslant \left|\sum_{n=1}^N[\phi_{nk}(x) - h_{n+1}(x) + h_n(x)] \right|+ \sum_{n= N+1}^\infty |\phi_{nk}(x)| + \sum_{n= N+1}^\infty |h_{n+1}(x)- h_n(x)| $$

The first sum on the RHS is finite and converges to $0$ as $k \to \infty$ since $\phi_{nk} \to h_{n+1} - h_n$. The second and third terms are tails of uniformly convergent series and can be made smaller than any $\epsilon > 0$ by choosing sufficiently large $N$.


$$\lim_{k \to \infty} \psi_k(x) = \sum_{n=1}^\infty [h_{n+1}(x) - h_n(x)],$$

and the sum on the RHS is Baire 1.


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.