Orbit equivalence of Cantor minimal systems Thanks: The first author acknowledges the partial support of his research by the National Natural Science Foundation of China (NSFC) grants 12271263 and 12250710128.Thanks: The second author acknowledges the partial support of his research by the National Natural Science Foundation of China (NSFC) grant 124B2001.

In this subsection we recall some basic notions and results in the descriptive set theory of equivalence relations. For other undefined notions and more details, the reader can consult [17] and [6].

A topological space $X$ is Polish if $X$ is separable and completely metrizable. A $G_{\delta}$ subset of a topological space is the intersection of countably many open sets. In a Polish space $X$, a subspace $Y$ is Polish if and only if $Y$ is a $G_{\delta}$ subset of $X$. A subset $Y$ of $X$ is comeager or generic if $Y$ contains a dense $G_{\delta}$ subset of $X$; a subset of $X$ is meager if its complement is comeager.

If $X$ is a Polish space and $\mathcal{B}$ is the $\sigma$-algebra generated by the open sets of $X$, then sets in $\mathcal{B}$ are called Borel sets. More abstractly, a Borel space is a pair $(X,\mathcal{B})$ where $X$ is a set and $\mathcal{B}$ is a $\sigma$-algebra of subsets of $X$; sets in $\mathcal{B}$ are again called Borel sets. A standard Borel space is a Borel space $(X,\mathcal{B})$ where there exists a Polish topology $\tau$ on $X$ such that $\mathcal{B}$ is the $\sigma$-algebra generated by $\tau$. We will use the following fact (see [17, Corollary 13.4]): If $(X,\mathcal{B})$ is a standard Borel space and $Y\in\mathcal{B}$, then $(Y,\mathcal{B}\upharpoonright Y)$ is again a standard Borel space, where

For notational simplicity, we write $X$ for a standard Borel space $(X,\mathcal{B})$ if there is no danger of confusion.

If $X,Y$ are Polish spaces and $f\colon X\to Y$ is a map, then we say that $f$ is continuous if for any open subset $U$ of $Y$, $f^{-1}(U)$ is an open subset of $X$. If $X,Y$ are Polish spaces or standard Borel spaces and $f\colon X\to Y$ is a map, then we say that $f$ is Borel (or Borel measurable) if for any Borel subset $B$ of $Y$, $f^{-1}(B)$ is a Borel subset of $X$. If $X$ is a Polish space or a standard Borel space, a subset $A$ of $X$ is analytic if there is a standard Borel space $Z$ and a Borel map $g\colon Z\to X$ such that $A=g(Z)$; $A$ is coanalytic if $X\setminus A$ is analytic.

An equivalence relation $E$ on $X$ is said to be Borel (or analytic) if $E$ is a Borel (or analytic, respectively) subset of $X\times X$. As we already recalled in the Introduction, if $E,F$ are equivalence relations on Polish or standard Borel spaces $X,Y$, respectively, then $E$ is Borel reducible to $F$, denoted $E\leq_{B}F$, if there is a Borel map $f\colon X\to Y$ such that for any $x_{1},x_{2}\in X$, we have

We say that $E$ is Borel bireducible with $F$, denoted $E\sim_{B}F$, if both $E\leq_{B}F$ and $F\leq_{B}E$. We say that $E$ is strictly Borel reducible to $F$, denoted $E<_{B}F$, if $E\leq_{B}F$ but $F\not\leq_{B}E$.

A topological group $G$ is Polish if its topology is Polish. If $G$ is a Polish group and $X$ is a Polish space, then an action $a\colon G\times X\to X$ of $G$ on $X$ is continuous if $a$ is a continuous map from the Polish space $G\times X$ to the Polish space $X$. If $G$ is a Polish space and $X$ is a Polish space or a standard Borel space, then an action $a\colon G\times X\to X$ is Borel if $a$ is a Borel map from $G\times X$ to $X$. For notational simplicity, we use $G\curvearrowright_{a}X$ or $G\curvearrowright X$ to denote the action $a$, and use $g\cdot x$ to denote $a(g,x)$.

If $G\curvearrowright X$ is a continuous or Borel action of a Polish group $G$ on a Polish or standard Borel space $X$, then for any $x\in X$, the orbit of $x$ is the set $G\cdot x=\mbox{\rm orbit}_{G}(x)=\{g\cdot x\colon g\in G\}$. The orbit equivalence relation induced by the action $G\curvearrowright X$ is defined as

In general, the equivalence relation $E^{X}_{G}$ is analytic and not necessarily Borel.

It is a theorem of Becker and Kechris (see, e.g. [6, Theorem 3.3.4]) that for any Polish group $G$ there exists a Borel action $G\curvearrowright X$, where $X$ is a standard Borel space, such that for any Borel action $G\curvearrowright Y$, where $Y$ is a standard Borel space, we have $E^{Y}_{G}\leq_{B}E^{X}_{G}$. The orbit equivalence relation $E^{X}_{G}$ is called a universal $G$-orbit equivalence relation, and is denoted $E^{\infty}_{G}$, or simply $E_{G}$. If $H$ is a closed subgroup of the Polish group $G$, then $H$ is also Polish and $E_{H}\leq_{B}E_{G}$.

A Polish group that is the most relevant to this paper is $S_{\infty}$, the group of all permutations of $\mathbb{N}$. Equipped with the subspace topology of $\mathbb{N}^{\mathbb{N}}$, $S_{\infty}$ is a Polish group. The universal $S_{\infty}$-orbit equivalence relation $E_{S_{\infty}}$ is one of the most well-studied benchmark equivalence relation in the descriptive set theory of equivalence relations. It is known to be analytic and not Borel, and Borel bireducible with many natural equivalence relations such as the isomorphism relation for all countable graphs, groups, fields, etc. (these are results due to Friedman and Stanley [5]; also see [6, Chapter 13]). In particular, the isomorphism of any Borel class of countable structures is Borel reducible to $E_{S_{\infty}}$. Thus, one direction of Melleray’s Theorem 1.2 (the Borel reduction to (v)) follows immediately from Theorem 1.1 (2).

The equivalence relation $=^{+}$ defined in Theorem 1.3 (viii) can easily seen to be an $S_{\infty}$-orbit equivalence relation. However, it is Borel, hence is strictly Borel reducible to $E_{S_{\infty}}$.

If the Polish group $G$ is a discrete countable group, then $E_{G}$ is always a Borel equivalence relation. Moreover, all of the orbits of $E_{G}$ is countable. Such equivalence relations are called countable; namely, a Borel equivalence relation $E$ is countable if every $E$-equivalence class is countable. Among all countable Borel equivalence relations, $E_{\mathbb{F}_{2}}$ is the universal one, i.e. $E\leq_{B}E_{\mathbb{F}_{2}}$ for any countable Borel equivalence relation, where $\mathbb{F}_{2}$ is the free group with two generators equipped with the discrete topology. This is a theorem of Jackson, Kechris and Louveau [15] (also see [6, Theorem 7.3.8]). It follows from results of Hjorth, Kechris and Louveau [14] that $E_{\mathbb{F}_{2}}<_{B}\ =^{+}$. Thus $E<_{B}\ =^{+}$ for any countable Borel equivalence relation.

Borel reducibility among countable Borel equivalence relations is very complicated. In this paper we only mention the following basic concepts. First, the simplest, as well as the least complex, Borel equivalence relation with uncountably many equivalence classes is the equality relation on $2^{\mathbb{N}}$, which we denote by $=$. An equivalence relation is smooth if it is Borel reducible to $=$. Thus smooth equivalence relations are the least complex in the Borel reducibility hierarchy among Borel equivalence relations with uncountably many equivalence classes. Among nonsmooth Borel equivalence relations, the least complex one is the equivalence relation $E_{0}$ on $2^{\mathbb{N}}$ defined as

This is a celebrated theorem of Harrington, Kechris and Louveau [11]. $E_{0}$ is clearly a countable Borel equivalence relation. We call a countable Borel equivalence relation hyperfinite if it is Borel reducible to $E_{0}$.

We have the following reducibility relations among the equivalence relations considered in this section:

Next we review the notion of amenability. A countable group $G$ is amenable if there is a left-invariant finitely additive probability measure on $G$. One can also define a notion of amenability for countable Borel equivalence relations as follows. Suppose that $\mu$ is a Borel measure on a standard Borel space $X$ and $E$ is a countable Borel equivalence relation on $X$. We say that $E$ is $\mu$-amenable if there is a $\mu$-measurable assignment $x\mapsto\nu_{x}$ such that:

• •

  each $\nu_{x}$ is a finitely additive probability measure on $X$ satisfying that $\nu_{x}([x]_{E})=1$, where $[x]_{E}$ is the $E$-equivalence class of $x$;

• •

  if $xEy$ then $\nu_{x}=\nu_{y}$.

Here the $\mu$-meaurability of the assignment $x\mapsto\nu_{x}$ means that for any Borel subset $B$ of $X\times X$, the map

is $\mu$-measurable. Now, a countable Borel equivalence relation $E$ on $X$ is amenable if it is $\mu$-amenable with respect to every Borel measure $\mu$ on $X$. If $G$ is an amenable countable group and $G\curvearrowright X$ is a Borel action, then $E^{X}_{G}$ is an amenable, countable Borel equivalence relation. It is known that $E_{\mathbb{F}_{2}}$ is not amenable (see, e.g., [6, Proposition 7.4.4 and Theorem 7.4.8]). Thus, if $E$ is a nonsmooth, amenable countable Borel equivalence relation, then $E_{0}\leq_{B}E<_{B}E_{\mathbb{F}_{2}}$.

Finally, a countable Borel equivalence relation $E$ on a standard Borel space $X$ is treeable if there is a Borel assignment $x\mapsto T_{x}$ such that

• •

  each $T_{x}$ is an acyclic graph (tree) with $[x]_{E}$ as the vertex set;

• •

  if $xEy$ then $T_{x}=T_{y}$.

Here the Borelness of the assignment $x\mapsto T_{x}$ means that the edge relation $\{(x,y,z)\in X^{3}\colon\{y,z\}\in T_{x}\}$ is Borel. Hyperfinite equivalence relations are treeable, and $E_{\mathbb{F}_{2}}$ is not treeable ([15, Corollary 3.28]). A recent result of Naryshkin and Vaccaro [20] states that if an orbit equivalence relation $E$ is induced by a Borel action of a countable amenable group, then $E$ is hyperfinite if and only if $E$ is treeable.

For a Bore equivalence relation $E$ which is not necessarily countable, and for a property $\mathsf{P}\in\{\mbox{countable},\mbox{treeable},\mbox{amenable},\mbox{hyperfinite}\}$, we say that $E$ is virtually $\mathsf{P}$ if $E$ is Borel reducible to a countable Borel equivalence relation with property $\mathsf{P}$. This is a new terminology which was previously called essentially $\mathsf{P}$ in the literature. Following [18, Section 4.1], we now say that $E$ is essentially $\mathsf{P}$ if $E$ is Borel bireducible with a countable Borel equivalence relation with property $\mathsf{P}$. It is known that essentially countable and virtually countable are different concepts ([13]). However, by the Harrington–Kechris–Louveau dichotomy theorem (see, e.g., [6, Theorem 6.3.1]), a Borel equivalence relation is virtually hyperfinite if and only if it is essentially hyperfinite.

In this subsection we establish some notation and basic results about the space of all Cantor minimal systems. Our treatment follows [8] unless otherwise indicated.

A Cantor set is a $0$-dimensional compact Polish space without isolated points. It is well-known that all Cantor sets are homeomorphic to $\mathcal{C}=2^{\mathbb{Z}}=\{0,1\}^{\mathbb{Z}}$. Let $\mbox{\rm Aut}(\mathcal{C})$ be the group of all autohomeomorphisms of $\mathcal{C}$ where the group operation is composition. Equipped with the supnorm metric, $\mbox{\rm Aut}(\mathcal{C})$ becomes a Polish group. Let $M(\mathcal{C})$ be the set of all minimal homeomorphisms in $\mbox{\rm Aut}(\mathcal{C})$. Then $M(\mathcal{C})$ is a $G_{\delta}$ subset of $\mbox{\rm Aut}(\mathcal{C})$, and hence is a Polish space itself. We view $M(\mathcal{C})$ as the space of all Cantor minimal systems. The orbit equivalence of Cantor minimal systems is thus an analytic equivalence relation on $M(\mathcal{C})$. It follows from Melleray’s Theorem 1.2 that this equivalence relation is not Borel.

The set of all odometers is a dense $G_{\delta}$ subset of $M(\mathcal{C})$ (see, e.g., [8, Proposition 3.6] or [19, Lemma 4.1]). More generally, for each positive integer $K$, the set of all Cantor minimal systems of topological rank at most $K$ is also a $G_{\delta}$ subset of $M(\mathcal{C})$ (see, e.g., [8, Corollary 3.4]). Thus, for any positive integer $K$, the set of all Cantor minimal systems of topological rank $K$ is a Borel subset of $M(\mathcal{C})$, and hence is a standard Borel space. It follows that the set of all Cantor minimal systems of finite topological rank is a standard Borel space as well.

By a theorem of Downarowicz and Maass [3], there are only two kinds of Cantor minimal systems of finite topological rank; when it has topological rank $1$, it is an odometer, and when it has topological rank $K>1$, it is topologically conjugate to a minimal subshift. We recall the definition of subshifts. Let $\mathsf{A}$ be a finite set (which we call the alphabet). The space $\mathsf{A}^{\mathbb{Z}}$ is a Cantor set. Consider the shift action $\mathbb{Z}\curvearrowright\mathsf{A}^{\mathbb{Z}}$: for any $g\in\mathbb{Z}$ and $x\in\mathsf{A}^{\mathbb{Z}}$,

for any $n\in\mathbb{Z}$. Then $(\mathsf{A}^{\mathbb{Z}},\varphi_{1})$ is a Cantor system. Given a Cantor system $(X,\varphi)$, a subset $A$ of $X$ is $\varphi$-invariant if $\varphi(A)=A$. A subshift $X$ is a closed $\varphi_{1}$-invariant subset of $\mathsf{A}^{\mathbb{Z}}$. If $X$ is a subshift, then $(X,\varphi_{1}\upharpoonright X)$ is a Cantor system.

Since the set of all odometers is a comeager subset of $M(\mathcal{C})$, the set of all minimal subshifts is thus a meager subset of $M(\mathcal{C})$. Following Pavlov and Schmieding [21], one can consider a suitable Polish space of subshifts which include all minimal subshifts (we do not give the details here since we do not use them in this paper). In this Polish space, the set of all minimal subshifts is a comeager Borel subset (see [21, Theorem 1.3]), hence it is possible to both consider it as a standard Borel space and speak of genericity among minimal subshifts. According to [21, Theorem 1.3], the set of all regular Toeplitz subshifts of topological rank $2$ is a comeager set of minimal subshifts. Each of the concepts in this statement corresponds to a Borel set. Moreover, it is shown in [21, Theorem 1.3] that a generic minimal subshift is orbit equivalent to the universal odometer. This is in contrast with our results (Theorems 1.4 and 1.5) that for any $K>1$, the orbit equivalence of minimal subshifts of topological rank $K$ is not smooth.

Given a dynamical systems $(X,\varphi)$, a Borel measure $\mu$ on $X$ is $\varphi$-invariant if $\mu(B)=\mu(\varphi^{-1}(B))$ for all Borel subsets $B$ of $X$. $\mu$ is ergodic if for every $\varphi$-invariant Borel set $B$, either $\mu(B)=0$ or $\mu(X\setminus B)=0$. The ergodicity of $\mu$ is equivalent to the following: if $\mu=a\mu_{1}+(1-a)\mu_{2}$ for two measures and some $0<a<1$, then $\mu=\mu_{1}=\mu_{2}$.

Given a Cantor space $X$, we denote by $\mbox{Prob}(X)$ the set of Borel probability measures on $X$. Note that a Borel probability measure on $X$ is uniquely determined by its values on clopen sets, and that any finitely additive probability measure on $\mbox{Clopen}(X)$ extends to an element of $\mbox{Prob}(X)$. Let $C(X)$ denotes the space of all real-valued continuous functions on $X$. Let $C(X)^{*}$ be the dual space of $C(X)$ equipped with the weak* topology. We can regard $\mbox{Prob}(X)$ as a subspace of $C(X)^{*}$, where the inherited topology on $\mbox{Prob}(X)$ is a natural compact topology induced by the maps $\mu\mapsto\mu(U)$, where $U$ ranges over all clopen subsets of $X$. Define a metric on $\mbox{Prob}(X)$ as

for any $\mu,\nu\in\mbox{Prob}(X)$. This is a compatible metric for the weak* topology, which makes $\mbox{Prob}(X)$ a compact Polish space.

Given a Cantor system $(X,\varphi)$, the set of all $\varphi$-invariant Borel probability measures on $X$ is nonempty, and we denote it by $\mbox{M}_{\varphi}$. Note that $\mbox{M}_{\varphi}$ is compact subset of $\mbox{Prob}(X)$. Let $\mbox{ME}_{\varphi}$ be the set of all ergodic measures on $(X,\varphi)$. Suppose that $(X,\varphi)$ is a Cantor minimal system with finitely many ergodic measures. Then $\mbox{EM}_{\varphi}$ is also a compact subset of $\mbox{Prob}(X)$. Let $\mu_{1},\dots,\mu_{n}$ be an enumeration of all of its ergodic measures. Then for every $\varphi$-invariant measure $\mu\in\mbox{Prob}(X)$, there are positive numbers $a_{1},\dots,a_{n}$ such that $a_{1}+\cdots+a_{n}=1$ and

Given a compact Polish space $Z$, we denote by $\mathcal{K}(Z)$ the space of all compact subsets of $Z$ equipped with the Vietoris topology. Then $\mathcal{K}(Z)$ is again a compact Polish space (see, e.g., [17, Theorem 4.26]).

In the next section we will consider the orbit equivalence of Cantor minimal systems with finitely many ergodic measures. In the following we show that it is a Borel class and that the map that assigns the set of ergodic measures to each such system is a Borel map.

We first recall some definitions and theorems from [10].

A countable abelian group $G$ together with a subset $G^{+}$ is called an ordered group if $G^{+}+G^{+}\subseteq G^{+}$, $G^{+}-G^{+}=G$ and $G^{+}\cap(-G^{+})=\varnothing$. Here $G^{+}$ is called the positive cone. We shall write $a\leq b$ if $b-a\in G^{+}$. If $(G,G^{+})$ is an ordered group and $H$ is a subgroup of $G$, then we set $H^{+}=H\cap G^{+}$. An element $u\in G^{+}$ is an order unit for $(G,G^{+})$ if for every $a\in G$, $a\leq nu$ for some $n\in\mathbb{N}$. We say that the ordered group $(G,G^{+})$ is unperforated if whenever $a\in G$ and $na\in G^{+}$ for some $n\in\mathbb{N}$, we have $a\in G^{+}$.

An unperforated ordered group $(G,G^{+})$ is a dimension group if given $a_{1}$, $a_{2}$, $b_{1}$, $b_{2}\in G$ with $a_{i}\leq b_{j}$ for all $i,j=1,2$, there exists a $c\in G$ with $a_{i}\leq c\leq b_{j}$ for all $i,j=1,2$. Let $(G_{1},G_{1}^{+},u_{1})$ and $(G_{2},G_{2}^{+},u_{2})$ be two dimension groups with distinguished order units. We say that they are isomorphic, and write $(G_{1},G_{1}^{+},u_{1})\cong(G_{2},G_{2}^{+},u_{2})$, if there exists a group isomorphism $f\colon G_{1}\to G_{2}$ such that $f(G_{1}^{+})=G_{2}^{+}$ and $f(u_{1})=u_{2}$.

If $(G,G^{+})$ is an ordered group, an order ideal is a subgroup $H$ of $G$ such that $H^{+}-H^{+}=H$ and whenever $0\leq a\leq b$ and $b\in H$, we have $a\in H$. A dimension group $(G,G^{+})$ is simple if $G$ does not contain any nontrivial order ideal. If $G$ is a simple dimension group it is easily seen that any $a\in G^{+}\setminus\{0\}$ is an order unit. Let $G$ be a simple dimension group and let $u\in G^{+}\setminus\{0\}$. We say that $a\in G$ is infinitesimal if for all positive integers $p,q$, we have $-pu\leq qa\leq pu$. Informally, we also write this condition as $-\epsilon u\leq a\leq\epsilon u$ for all $0<\epsilon\in\mathbb{Q}$. The collection of infinitesimal elements of $G$ form a subgroup, the infinitesimal subgroup of $G$, which we denote by $\mbox{Inf}(G)$.

The quotient group $G/\mbox{Inf}(G)$ has a natural induced ordering. Let $a\mapsto\hat{a}$ denote the quotient map from $G$ onto $G/\mbox{Inf}(G)$. Then we can set $\hat{a}>0$ iff $a>0$. It is easily seen that $G/\mbox{Inf}(G)$ is a simple dimension group with no infinitesimal elements except 0. If $G$ has distinguished order unit $u$ then $G/\mbox{Inf}(G)$ inherits the distinguished order unit $\hat{u}$.

Let $(X,\varphi)$ be a Cantor minimal system. Denote by $C(X,\mathbb{Z})$ the set of the continuous maps from $X$ to $\mathbb{Z}$. So $C(X,\mathbb{Z})$ is a countable abelian group under addition. Let $B_{\varphi}$ denote the coboundary subgroup $\{f-f\circ\varphi^{-1}\colon f\in C(X,\mathbb{Z})\}$. Define $K^{0}(X,\varphi)$ as the quotient group $C(X,\mathbb{Z})/B_{\varphi}$ and let

where $f\mapsto[f]$ denotes the quotient map. Let $\mathbf{1}$ denote the element of $K^{0}(X,\varphi)^{+}$ that the constant function $1$ maps to. Then $K^{0}(X,\varphi)$ with the positive cone $K^{0}(X,\varphi)^{+}$ is a simple, torsion-free dimension group with distinguished order unit $\hat{\mathbf{1}}$.

We will use the following result, which was explicitly stated as a part of [10, Theorem 1.13] but can be traced back to [12, Theorem 5.5] and [4, Corollary 4.2].

Suppose that $(X,\varphi)$ is a Cantor minimal system with finitely many ergodic measures. Let $\mu_{1},\dots,\mu_{n}$ be a nonrepetitive enumeration of its ergodic measures. Define $\Gamma_{\varphi}$ as

It is clear that $\Gamma_{\varphi}$ is a subgroup of $\mathbb{R}^{n}$ and

is a dimension group with distinguished order unit, where $1^{n}=(1,\dots,1)\in\mathbb{R}^{n}$.

We first note the following direct corollary of Theorem 1.1.

If $G$ is a subgroup of $\mathbb{R}^{n}$ and $\phi\in\mbox{\rm Sym}(n)$, we define

Then $\phi(G)$ is still a subgroup of $\mathbb{R}^{n}$. The following is a characterization of orbit equivalence for Cantor minimal systems with finitely many ergodic measures.

The above lemma demonstrates that the orbit equivalence for Cantor minimal systems is a Borel equivalence relation. In the following, we determine its complexity in the Borel reducibility hierarchy.

This theorem gives the upper bound for the complexity of the orbit equivalence for Cantor minimal systems with finitely many ergodic measures.

We consider a special kind of Cantor minimal systems, namely Toeplitz subshifts.

We first recall some basic definitions and results about Toeplitz subshifts following [23].

Let $\mathsf{A}$ be a finite set. A subshift $X\subseteq\mathsf{A}^{\mathbb{Z}}$ is Toeplitz if $X$ is the closure of $\mbox{\rm orbit}_{\varphi_{1}}(x)$ for some $x\in\mathsf{A}^{\mathbb{Z}}$, where $x$ satisfies that for any $k\in\mathbb{Z}$ there is $p\in\mathbb{N}$ such that $x(k)=x(k+np)$ for all $n\in\mathbb{Z}$. Such an $x$ is called a Toeplitz sequence; and we say that $X$ is generated by $x$. In the context of subshifts we usually denote the shift map $\varphi_{1}$ as $\sigma$. When we want to emphasize the alphabet, we also call $x$ an $\mathsf{A}$-Toeplitz sequence and $X$ an $\mathsf{A}$-Toeplitz subshift.

Let $x$ be an $\mathsf{A}$-Toeplitz sequence. For $p\in\mathbb{N}$, let

Fix a new symbol $\Box\not\in\mathsf{A}$. Define the $p$-skeleton of $x$ as an element $x^{\prime}\in(A\cup\{\Box\})^{\mathbb{Z}}$ where

Then the $p$-skeleton of $x$ is periodic with period $p$. We call $p$ an essential period of $x$ if the $p$-skeleton of $x$ is not periodic with any smaller period. Let $P(x)$ denote the set of all essential periods of $x$. For $p,q\in P(x)$, let $p\sqsubseteq q$ if $p$ is a factor of $q$. Then $(P(x),\sqsubseteq)$ is a directed set (see [23, Proposition 2.1]). For each $p\in P(x)$, define