---
title: Disintegration Theorem for Markov Kernels
url: https://www.emergentmind.com/topics/disintegration-theorem-for-markov-kernels
type: topic
---

# Disintegration Theorem for Markov Kernels

The disintegration theorem for Markov kernels is the statement that a joint measure or joint kernel can be decomposed into a marginal object and a conditional kernel. In kernel notation, for a finite kernel $\kappa : \mathcal X \rightsquigarrow \mathcal Y \times \mathcal Z$, one seeks a kernel $\eta : \mathcal X \times \mathcal Y \rightsquigarrow \mathcal Z$ such that $\kappa = \kappa_{\mathrm{fst}} \otimes \eta$; in the equivalent measure-theoretic formulation, a measure $\mu$ together with a measurable map $\pi : X \to Y$ is represented as $\mu = \nu \otimes \mu_y$ with $\nu = \pi_*\mu$ and the conditional measures $\mu_y$ concentrated on the fibers $\pi^{-1}(y)$ [2510.04070]. In contemporary work, this theorem underlies regular conditional distributions, posterior kernels, conditional expectation kernels, formalized probability theory, and several geometric and computability-theoretic refinements of conditioning [2202.04511].

## 1. Kernel-theoretic formulation

A Markov kernel is a measurable family of probability measures. In the Mathlib formalization, a kernel from a measurable space $\mathcal X$ to a measurable space $\mathcal Y$ is literally a measurable function
\[
\kappa : \mathcal X \to \mathcal M(\mathcal Y),
\]
where measurability means that for every measurable set $B \subseteq \mathcal Y$, the map $x \mapsto \kappa(x)(B)$ is measurable; a Markov kernel is then a kernel for which each $\kappa(x)$ is a probability measure [2510.04070].

Within that language, disintegration is the kernel-level version of the probabilistic factorization
\[
\text{joint law} = \text{marginal} \otimes \text{conditional law}.
\]
The formal theorem described in Mathlib states that if $\mathcal Z$ is a standard Borel space and either $\mathcal X$ is countable or $\mathcal Y$ has a countably generated $\sigma$-algebra, then every finite kernel
\[
\kappa : \mathcal X \rightsquigarrow \mathcal Y \times \mathcal Z
\]
admits a disintegration. Informally, there exists a conditional kernel
\[
\eta : \mathcal X \times \mathcal Y \rightsquigarrow \mathcal Z
\]
such that $\kappa = \kappa_{\mathrm{fst}} \otimes \eta$ [2510.04070].

This formulation packages conditioning as an equality of kernels rather than as an almost-everywhere identity for random variables. A plausible implication is that the theorem is best understood not merely as a construction of regular conditional probabilities, but as a structural factorization principle for stochastic morphisms.

## 2. Measure disintegration, fibers, and regular conditional laws

A classical measure-space version is given in the transport-geometric setting of locally compact and separable metric spaces. If $X$ and $Y$ are locally compact and separable metric spaces, $\pi : X \to Y$ is a Borel map, and $\mu \in \mathcal M_{+}(X)$ is a positive finite Radon measure with $\nu := \pi_*\mu$, then there exist measures $\mu_y \in \mathcal M_{+}(X)$ such that $y \mapsto \mu_y$ is a Borel map, $\mu_y \in \mathscr P(X)$ for $\nu$-a.e. $y$, the measure $\mu$ disintegrates as
\[
\mu = \nu \otimes \mu_y,
\qquad
\mu(A)= \int_Y \mu_y(A)\, d\nu(y),
\]
and $\mu_y$ is concentrated on the fiber $\pi^{-1}(y)$ for $\nu$-a.e. $y$ [2202.04511].

The equivalent integral form is
\[
\int_Y \int_{\pi^{-1}(y)} \omega(x)\, d\mu_y\, d\nu(y) = \int_X \omega(x)\, d\mu(x)
\]
for $\omega \in C_c(X)$, and then by extension for Borel sets and Borel functions [2202.04511]. In this setting, the family $y \mapsto \mu_y$ is exactly a regular conditional probability kernel: for each $y$, $A \mapsto \mu_y(A)$ is a probability measure; for each Borel set $A$, the map $y \mapsto \mu_y(A)$ is measurable; and the original measure is recovered by integration against $\nu$ [2202.04511].

A 2011 abstract-level contribution described a new approach to disintegration of measures that allows one to drop the usually taken separability assumption, using a result on fibers in the spectrum of algebra of essentially bounded functions as its main tool [1103.0255]. Since only the abstract-level statement is available there, the precise kernel-theoretic form is not specified in that source summary. This suggests a nonseparable extension of the standard separable or standard Borel paradigms emphasized in later accounts.

## 3. Existence hypotheses, uniqueness, and canonicality

The cited literature does not treat arbitrary measurable spaces uniformly. In the Mathlib formalization, the disintegration theorem is stated under the hypotheses that $\mathcal Z$ is standard Borel and that either $\mathcal X$ is countable or $\mathcal Y$ has a countably generated $\sigma$-algebra; these are described as exactly the assumptions that make the measurable selection and density construction work in the formal proof [2510.04070]. In the transport-geometric theorem, the ambient spaces are locally compact and separable metric spaces and the base measure is a positive finite Radon measure [2202.04511]. In the computable-analysis treatment, the setting is complete separable metric spaces along a projection map [1509.02992].

Uniqueness is typically only almost-everywhere. The Mathlib account stresses that conditional kernels are almost everywhere equal to a Markov kernel and almost everywhere unique; moreover, if $\eta$ disintegrates $\kappa$ and $\eta'$ is equal to $\eta$ almost everywhere with respect to $\kappa_{\mathrm{fst}}$, then $\eta'$ also disintegrates $\kappa$ [2510.04070]. In the geometric formulation, a uniqueness proposition states that if
\[
\gamma=\nu\otimes \eta_x,\qquad \eta_x \text{ concentrated on } \operatorname{proj}_X^{-1}(x),
\]
then the family $\eta_x$ is unique $\nu$-a.e. [2202.04511].

A stronger form of canonicity appears when disintegrations are continuous. On the domain of measures admitting a continuous disintegration along the projection $\pi_2 : S \times T \to T$ and whose marginal $\#1{\pi_2}\mu$ has full support, the disintegration is necessarily unique. The same work relates this to Tjur’s pointwise notion of conditioning: under mild assumptions, the existence of a continuous disintegration on a measure-one set is equivalent to every point in that set being a Tjur point [1509.02992].

When continuous disintegration fails, pointwise conditioning can still be recovered from a Vitali basis. The Fraser–Naderi construction defines
\[
\mu_{g,V}^x(A) :=\lim_{n\to\infty} \frac{\mu(g^{-1}(E_n)\cap A)}{\mu(g^{-1}(E_n))}
\]
for sequences $E_n$ converging regularly to $x$, and yields a measurable disintegration satisfying the correct integral identity. However, the resulting pointwise version may depend on the chosen class $V$ [1509.02992]. A common misconception is that disintegration always yields a canonical pointwise conditional law; the literature instead separates almost-everywhere uniqueness, continuous uniqueness, and basis-dependent pointwise realizations.

## 4. Formalization in proof assistants and proof architecture

The theorem has been formalized in Lean within Mathlib’s probability library, specifically in
\[
\texttt{Mathlib/Probability/Kernel/Disintegration/StandardBorel.lean},
\]
where disintegration is expressed as an equality between kernels rather than as an external existence statement [2510.04070]. The library treats kernels as first-class objects, and this design is presented as central rather than auxiliary.

The formal proof proceeds through several intermediate kernel constructions. First, densities for finite kernels are built in a jointly measurable form,
\[
d_{\kappa,\nu}(x,y)(C),
\]
so that
\[
\int_{y\in B} d_{\kappa,\nu}(x,y)(C)\,d\nu(x)=\kappa(x)(B\times C).
\]
Second, a Radon–Nikodym theorem for kernels constructs a jointly measurable derivative
\[
\frac{d\kappa}{d\eta}
\]
together with a singular remainder. Third, for $\mathcal Z = \mathbb R$, a conditional CDF is built and converted into a kernel using Stieltjes measures. Fourth, the standard Borel case is reduced to the real-valued case by measurable embedding into $\mathbb R$ [2510.04070].

This formalization is not isolated from downstream probability theory. It is used to define conditional distributions of random variables, posterior distributions, and conditional expectation kernels. For random variables $X : \Omega \to \mathcal X$ and $Y : \Omega \to \mathcal Y$ with $\mathcal Y$ standard Borel, disintegration yields a kernel $P_{Y|X} : \mathcal X \rightsquigarrow \mathcal Y$ such that
\[
P_{XY} = P_X \otimes_m P_{Y|X}.
\]
Given a prior measure $\mu$ on $\mathcal X$ and a kernel $\kappa : \mathcal X \rightsquigarrow \mathcal Y$, disintegrating the swapped joint law defines a posterior kernel $\kappa^\dagger_\mu : \mathcal Y \rightsquigarrow \mathcal X$ satisfying
\[
s_*(\mu \otimes_m \kappa) = (\kappa \circ_m \mu) \otimes_m \kappa^\dagger_\mu
\]
[2510.04070].

## 5. Geometric and computational perspectives

One contemporary line of work interprets disintegration through optimal transport. For a transport plan $\gamma \in \mathscr P(X \times Y)$ with marginals $\mu$ and $\nu$, one may disintegrate
\[
\gamma = \mu \otimes \gamma_x,
\]
and regard the conditional family as a disintegration map
\[
f : Y \to (\mathscr P(X), W_p), \qquad y \mapsto \mu_y,
\]
or, in product settings,
\[
f : X \to (\mathscr P(Y), W_1), \qquad x \mapsto \gamma_x.
\]
In the Monge case, if $T:X\to Y$ is a transport map, the disintegration map is $x \mapsto \delta_{T(x)}$ [2202.04511].

This geometric perspective yields regularity statements beyond bare existence. If $\nu \ll \operatorname{vol}_Y$, then the disintegration map is nearly weakly continuous; if $\pi$ is bijective and continuous, the conditional measures become Dirac masses,
\[
\mu_y = \delta_{\pi^{-1}(y)},
\]
and the map is weakly continuous; in the metric measure foliation setting, the identity
\[
W_2(\mu_y,\mu_{y'}) = d^*(y,y')
\]
makes the disintegration map an isometry, hence weakly continuous [2202.04511]. The same paper proves a rigidity theorem asserting that absolute continuity can propagate along Wasserstein paths of fiber measures when the disintegration map is weakly continuous and the stated geometric hypotheses hold [2202.04511].

A different refinement studies disintegration as a computational operator. On the domain of measures with a unique continuous disintegration, the disintegration operator
\[
\DD
\]
is strongly Weihrauch equivalent to the limit operator,
\[
\DD \equiv_{\mathrm{sW}} \Lim.
\]
The lower bound is obtained by encoding an arbitrary element of $C(\mathbb N,\mathbb S)$ into a computable measure whose unique continuous disintegration reveals that element, while the upper bound computes a continuous disintegration via a single limit procedure using a $\mu$-continuous basis and the Tjur property [1509.02992]. In the noncontinuous case, if a Vitali basis is computable, the induced pointwise disintegration remains strongly Weihrauch reducible to $\Lim$ [1509.02992]. This establishes that conditioning is, in general, not computable, even when existence is secured.

## 6. Related kernel constructions and conceptual boundaries

Disintegration also appears in higher-order kernel constructions where the conditioned objects are themselves kernels. One formulation defines the conditional distribution of a Markov kernel $M_1$ given another kernel $M_2$ as a kernel
\[
L : (\Omega_2,\mathcal A_2)\to(\Omega_1,\mathcal A_1)
\]
such that
\[
\int_\Omega M_1(\omega,A_1)M_2(\omega,A_2)\,dP(\omega)
=
\int_{\Omega_2} L(\omega_2,A_1)\,dP_{M_2}(\omega_2)
\]
for every $A_1 \in \mathcal A_1$ and $A_2 \in \mathcal A_2$. Existence is stated under standard regularity assumptions, for example when the relevant measurable spaces are standard Borel [1706.03955]. In that framework, conditional independence of kernels is defined by factorization of a conditional joint kernel, and a main theorem states that under $M_1 \perp_P M_2 \mid M_3$, unconditional independence of $M_1$ and $M_2$ is equivalent to independence of the conditional kernels $P_{M_1\mid M_3}$ and $P_{M_2\mid M_3}$ [1706.03955].

A related but narrower development studies conditional expectation for Markov kernels. There the conditional expectation of an integrable kernel $M_1$ given another kernel $M_2$ is defined via the Radon–Nikodym derivative
\[
E_P(M_1\mid M_2) :=\frac{d(M_1\cdot P)^{M_2}}{dP^{M_2}},
\]
and the main theorem proves the invariance identity
\[
E(M\mid M_1\times M_2)=E(M\mid M_1)
\]
under the independence assumption
\[
M_2 \perp_P (M\times M_1).
\]
That work explicitly does not claim a general disintegration theorem with existence and uniqueness of regular conditional probabilities; rather, it proves a conditional-expectation identity inside a kernel framework [2006.03328].

Disintegration-style reasoning also appears in the theory of invariant measures for Markov kernels. A recent uniqueness result on standard Borel spaces reduces multiplicity of invariant probability measures to a measurable decomposition into disjoint absorbing sets, using the mutual singularity of distinct invariant ergodic measures and an ergodic-decomposition argument [2601.04900]. This suggests that disintegration is not only a theorem about conditional laws, but also a recurrent structural pattern in the analysis of kernels: measures are decomposed into components, and qualitative properties are recovered from the behavior of those components.

Across these developments, the central content of the disintegration theorem remains stable. It asserts that a joint probabilistic object can be reconstructed from a marginal and a conditional kernel, but the surrounding theory varies according to the regularity sought: almost-everywhere existence, continuous versions, formalized equality of kernels, Wasserstein continuity, posterior inversion, or computability bounds.

Source: https://www.emergentmind.com/topics/disintegration-theorem-for-markov-kernels