I was reviewing Fubini theorem for general measures and noticed that I need $\{x\} \in \mathcal{F}_{X}$ to complete the proof:

Do I need $\{ x\} \in \mathcal{F}_X $ for all $x$ in order to state Fubini-(Toneli) theorem?

General Fubini theorem states that given $\sigma$-finite measure space $(X,\mathcal{F}_X,\mu)$, measurable space $(Y,\mathcal{F}_Y)$, a family of measures $\nu_x$ over $(Y,\mathcal{F}_Y)$ parametrized by elements of $X$ satisfying certain conditions, and measurable function $f$ over $(X \times Y,\mathcal{F}_X \times \mathcal{F}_Y)$ which is nonnegative (or has integral or integrable) the function of $x \in X$ $$ \int_Y f(x,y) \, d\nu_x(y) $$ is measurable, and

$$ \int_X \int_Y f(x,y) \, d \nu_x(y) \, d\mu(x) = \int_{X \times Y}f \, d \nu' $$ where $\nu'$ is uniquely defined by $ \nu_x$ and $\mu$

However, to make this theorem make sense we need to show for all $x$ that $f(x,\cdot)$ is measurable with respect to $\mathcal{F}_Y$. At first this seems trivial. Take Borel set $A \subset \mathbb{R}$. Then

$$(f(x, \cdot))^{-1}(A) = \pi_Y(f^{-1}(A) \cap \{x\} \times Y)$$

where $\pi_Y$ is a projection $\pi_Y : (a,b) \mapsto b$.

$f^{-1}(A)$ is measurable set by definition of $f$ and if $\{x\} \in \mathcal{F}_X$ the whole intersection will be measurable which is the case then $X = \mathbb{R}^n$ wiht Borel $\sigma$-algebra.

However intersection may happen to be non-measurable then $\{x\} \not \in \mathcal{F}_X$. So what to do in this case?

Call measurable space $(X, \mathcal{F}_X)$ such that $\{x\} \in \mathcal{F}_X$ for all $x \in X$ fubinible . Do I need to ask for $(X,\mathcal{F}_X)$ to be fubinible is statement of Fubini theorem? Are there other version of Fubini theorem which don't requere fubinible measurable space, say for point-free lattices?


The product $\sigma$-algebra $\mathcal{F}_X \times \mathcal{F}_Y$ has the property that all sections of measurable sets belong to the corresponding factor $\sigma$-algebra. That is, we have

$$A_x := \{ y \in Y : (x,y) \in A\} \in \mathcal{F}_Y\tag{$\ast$}$$

for all $A \in \mathcal{F}_X \times \mathcal{F}_Y$ and all $x\in X$, and the analogous statement with the roles of $X$ and $Y$ exchanged.

From $(\ast)$ it easily follows that each section $f(x,\,\cdot\,)$ is $\mathcal{F}_Y$-measurable for $\mathcal{F}_X\times \mathcal{F}_Y$-measurable $f$, since $\bigl(f(x,\,\cdot\,)\bigr)^{-1}(A) = \bigl(f^{-1}(A)\bigr)_x \in \mathcal{F}_Y$ for Borel sets $A$ by $(\ast)$.

Thus let's prove $(\ast)$. For every $x \in X$, define

$$\mathcal{S}_x := \{ A \subset X\times Y : A_x \in \mathcal{F}_Y\}.$$

Then since $\varnothing_x = \varnothing$, $(X\times Y\setminus A)_x = Y \setminus A_x$ and $\bigl(\bigcup A_n\bigr)_x = \bigcup (A_n)_x$, it follows that $\mathcal{S}_x$ is a $\sigma$-algebra. Also, we have $U\times V \in \mathcal{S}_x$ for all $U \in \mathcal{F}_X$ and $V\in \mathcal{F}_Y$, since $(U\times V)_x$ is either $\varnothing$ or $V$. By definition,

$$\mathcal{F}_X \times \mathcal{F}_Y = \sigma\bigl(\{ U \times V : U \in \mathcal{F}_X,\, V \in \mathcal{F}_Y\}\bigr)$$

is the smallest $\sigma$-algebra containing all measurable rectangles, so

$$\mathcal{F}_X \times \mathcal{F}_Y \subset \mathcal{S}_x.$$

But $x\in X$ was completely arbitrary, so

$$\mathcal{F}_X \times \mathcal{F}_Y \subset \bigcap_{x\in X} \mathcal{S}_x,$$

which is just $(\ast)$.

  • $\begingroup$ Thank you Daniel Fischer! At first I thought that proof will require axiom of choice but now as I see that inverse is just a section everything is clear. $\endgroup$ – Nik Pronko Nov 23 '16 at 16:48

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.