MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

According to Kolmogorov ("Algèbres de Boole métriques complètes", VI Zjazd Matematykòw Polskich, 1948, english translation Phil. Studies, 1995, 77, 57-66), a Boolean algebra $(B, \wedge, \vee,-,1,0)$ is metric if it is given together with a map $ B\xrightarrow\mu R $ to the reals such that:

1) $[x \wedge y=0] \Rightarrow [\mu(x \vee y)=\mu (x) + \mu (y)]$

2) $ [x \neq 0] \Rightarrow [\mu (x)>0] $.

In facts, if $B \times B \xrightarrow+ B$ denotes symmetric difference, the composite $ B \times B \xrightarrow+ B \xrightarrow\mu R$ is a metric.

Denoting: Bool the category of Boolean algebras, Met the category of metric spaces with non-expansive maps and MBool the category of metric Boolean algebras with non-expansive Boolean homomorphisms:

  • what are the main differences between the categories MBool and Bool ?

  • the two forgetful functors MBool**$\rightarrow$ *Bool and MBool*$\rightarrow$ **Met have left adjoints ?

share|cite|improve this question

I have a feeling that uncountable products in $MBool$ do not exist in general. One thing is for sure: if $MBool$ has uncountable products, then the forgetful functor $MBool \to Bool$ cannot preserve them (whence the forgetful functor couldn't have a left adjoint.)

Indeed, if $2$ is the 2-element Boolean algebra with its unique measure $\mu$. If the uncountable product $(2, \mu)^{\omega_1}$ existed in $MBool$, where $\omega_1$ is the first uncountable ordinal, and if this product is calculated as it would be in $Bool$ (which is the set-theoretic infinite product), then we could construct an uncountable strictly decreasing chain of elements (uncountable tuples)

$$(1, 1, 1, \ldots) > (0, 1, 1, \ldots) > (0, 0, 1, \ldots) > \ldots$$

and there would have to be a corresponding uncountable strictly decreasing chain of positive real numbers

$$\mu(1, 1, 1, \ldots) > \mu(0, 1, 1, \ldots) > \mu(0, 0, 1, \ldots) > \ldots$$

by the axioms on metric Boolean algebras. This is impossible by cofinality considerations.

Regardless of whether there are infinite products in $MBool$, there is a trivial reason why the forgetful functor $U: MBool \to Bool$ cannot have a left adjoint $F$ under the axioms given for metric Boolean algebras. Suppose WLOG that $\mu(1) = 1$ where the $1$ on the left is the top element of the putative free metric Boolean algebra $FB$. Then define a different metric $\mu'$ on $UFB$ by $\mu'(b) = r\mu(b)$ where $r > 1$, giving a different metric Boolean algebra $B'$. Then there is no nonexpansive metric Boolean algebra map $FB \to B'$ which extends the Boolean algebra embedding $i: B \to UB'$ along $i: B \to UFB$.

One might think this trivial objection could be remedied by adding an extra axiom like $\mu(1) = 1$ (so we are working with probability measures in a sense), but it seems highly doubtful even in that case that a left adjoint out of $Bool$ would exist. As noted above, it could only exist if $MBool$ lacked many limits like uncountable products, and this would rule out a straightforward application of an adjoint functor theorem.

I am not sure about a left adjoint out of $Met$.

share|cite|improve this answer
Thanks. Very useful insight. – Davide Bernardini Oct 18 '12 at 14:32

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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