---
title: 'MonoRel: Monotone Relations in CQ Containment'
url: https://www.emergentmind.com/topics/monotone-relations-monorel
type: topic
---

# MonoRel: Monotone Relations in CQ Containment

Searching arXiv for the primary paper and closely related work on monotone relations / MonoRel.
arXiv search query: 1808.08822
MonoRel denotes the monotone fragment of Boolean queries expressed as containments of conjunctive queries over a fixed finite relational schema $\mathcal T$. Formally, if $\mathrm{CQ}^{\subseteq}$ is the class of Boolean queries of the form $Q_1 \subseteq Q_2$ and $\mathrm{MON}$ is the class of monotone Boolean queries, then
\[
\mathrm{MonoRel}:=\mathrm{CQ}^{\subseteq}\cap \mathrm{MON}.
\]
Its significance lies in a preservation theorem showing that monotone containments of conjunctive queries add no expressive power beyond ordinary conjunctive-query nonemptiness: for every schema $\mathcal T$, $\mathrm{MonoRel}=\mathrm{CQ}^{\neq}$, where $\mathrm{CQ}^{\neq}$ is the class of Boolean queries given by nonemptiness of a single conjunctive query [1808.08822].

## 1. Formal setting and definition

The underlying framework fixes a finite relational schema $\mathcal T$, where each relation symbol $R\in\mathcal T$ has arity $\mathrm{arity}(R)$. An instance $I$ of $\mathcal T$ is a finite set of ground facts $R(a_1,\ldots,a_n)$, and $I\subseteq J$ means that $J$ extends $I$ by set-theoretic inclusion [1808.08822].

A conjunctive query $Q$ over $\mathcal T$ is written as
\[
Q:\ H \leftarrow B,
\]
where $H$ is the head, a tuple of variables $(x_1,\ldots,x_k)$, and $B$ is the body, a finite set of atoms $R(v_1,\ldots,v_n)$ with variables from $\mathrm{Var}$. The answer $Q(I)$ is the set of tuples $t=(t_1,\ldots,t_k)\in \mathrm{adom}(I)^k$ for which there exists a homomorphism $f:\mathrm{Var}\to \mathrm{adom}(I)$ satisfying the body atoms and $t_i=f(H(i))$ for all $i$. When the head is empty, $Q(I)$ is either $\{()\}$ or $\emptyset$ [1808.08822].

Within this framework, Boolean queries arise in two standard ways. The first is nonemptiness: a conjunctive query $Q$ induces the Boolean test $Q\neq\emptyset$, which returns true exactly when $Q(I)\neq\emptyset$. The second is containment: if $Q_1$ and $Q_2$ have the same head arity, then $Q_1\subseteq Q_2$ is the Boolean query that returns true on $I$ iff $Q_1(I)\subseteq Q_2(I)$. Monotonicity is defined semantically: a Boolean query $B$ is monotone if for all instances $I\subseteq J$,
\[
B(I)=\mathrm{true}\ \Rightarrow\ B(J)=\mathrm{true}.
\]
MonoRel is precisely the subcollection of containment queries with this monotonicity property [1808.08822].

## 2. Preservation theorem and expressive equivalence

The central theorem states that restricting containment-of-CQ queries to the monotone fragment yields no Boolean queries beyond conjunctive-query nonemptiness. In its global form,
\[
\mathrm{MonoRel}=\mathrm{CQ}^{\neq}.
\]
Equivalently, for every pair of conjunctive queries $Q_1,Q_2$, the Boolean query
\[
I\mapsto [Q_1(I)\subseteq Q_2(I)]
\]
is monotone if and only if there exists a conjunctive query $Q$ such that, for all instances $I$,
\[
Q_1(I)\subseteq Q_2(I)\iff Q(I)\neq\emptyset.
\]
The result is therefore a preservation theorem in the precise sense that a semantic restriction—monotonicity—forces a syntactic collapse to the language of CQ nonemptiness [1808.08822].

The theorem is asymmetric in difficulty. The easy direction is that every CQ nonemptiness query is monotone and can be represented within the containment formalism. The substantive content is the converse: given a monotone containment $Q_1\subseteq Q_2$, one can construct a single conjunctive query $Q$ whose nonemptiness reproduces exactly the same yes/no behavior on all finite instances [1808.08822].

This characterization isolates the exact expressive boundary of monotone CQ containments. Containment syntax may appear richer than nonemptiness syntax because it compares two query answers rather than inspecting one. The theorem shows that, once monotonicity is imposed, this apparent extra expressive latitude disappears [1808.08822].

## 3. Proof architecture

The proof proceeds through a structural reduction of the containment query. The first step is **Head Elimination**. If a monotone containment $Q_1\subseteq Q_2$ is given, then it is equivalent to one in which both heads are empty:
\[
()\leftarrow B_1 \ \subseteq\ ()\leftarrow B_2.
\]
The proof idea uses a one-element instance $Z_a$ containing a single fact $R(a,\ldots,a)$ for each $R\in\mathcal T$ to force homomorphisms to identify head positions, and then applies monotonicity to transfer the argument from $Z_a$ to arbitrary instances [1808.08822].

The second step is a minimality analysis of the left-hand body. If $Q_1=()\leftarrow B_1$ and $Q_2=()\leftarrow B_2$ with $B_1$ minimal, then a nontrivial monotone case can only occur when $|B_1|$ is very small. More precisely, if $|B_1|\ge 2$, then either the containment is equivalent to the constant true query or it fails to be monotone. The proof uses the Chandra–Merlin characterization of CQ containment together with the additivity of connected bodies to build a counterexample whenever the containment is nontrivial but not syntactically forced [1808.08822].

After these reductions, the proof turns to a case analysis on $|B_1|$. The one-atom case is decisive. If $B_1$ contains exactly one atom $R(\ldots)$, then the connected components of $B_2$ can be partitioned according to whether they must map into that single $R$-atom or into other parts of an instance. This yields a subset $C$ of the connected components of $B_2$ such that the original monotone containment is equivalent to the Boolean query defined by the conjunctive query
\[
Q:()\leftarrow C.
\]
Since nonemptiness of that $Q$ is a CQ nonemptiness test, the equivalence with $\mathrm{CQ}^{\neq}$ follows [1808.08822].

A plausible implication is that MonoRel is best viewed not as a genuinely new query formalism, but as a semantic subclass whose members admit canonical reduction to ordinary existential pattern detection.

## 4. Canonical example

A standard example uses the schema $\mathcal T=\{R(1),S(1)\}$ and the Boolean conjunctive queries
\[
Q_1=()\leftarrow R(x), \qquad Q_2=()\leftarrow R(x),S(y).
\]
For an instance $I$, the containment $Q_1\subseteq Q_2$ holds exactly when every $R$-fact in $I$ has a corresponding $S$-fact in $I$, that is:
\[
\text{if }R(x)\in I\text{ for some }x,\ \text{then }S(y)\in I\text{ for some }y.
\]
This containment is monotone, because adding facts can only make the implication easier to satisfy [1808.08822].

The preservation theorem requires a single CQ whose nonemptiness captures the same Boolean behavior. In this case, the body of $Q_2$ has two connected components, $\{R(x)\}$ and $\{S(y)\}$, and the monotone analysis selects the $S$-component. The resulting conjunctive query is
\[
Q:()\leftarrow S(y).
\]
Its nonemptiness test says that the instance contains at least one $S$-fact, which is exactly the condition that renders the presence of $R$-facts harmless with respect to the containment $Q_1\subseteq Q_2$ [1808.08822].

This example exhibits the general mechanism of the theorem. The containment query is not interpreted as a comparison between two arbitrary answer sets; instead, its monotone content is distilled into the existence of a specific component pattern already present in the right-hand query.

## 5. Consequences for syntax, expressivity, and complexity

The theorem yields a **syntactic characterization** of a semantic notion. The condition “$Q_1\subseteq Q_2$ is monotone” is semantic, since it quantifies over all instance extensions. The result shows that this semantic condition is captured exactly by the syntactic language $\mathrm{CQ}^{\neq}$ [1808.08822].

A second consequence is **no extra power**. Allowing containment of two conjunctive queries may appear to enlarge the Boolean expressive resources available to database query languages. For monotone Boolean queries, however, it does not: beyond testing CQ nonemptiness, containment syntax contributes no additional expressive power [1808.08822].

A third consequence concerns **complexity**. As a by-product of the same case distinctions used in the proof, testing monotonicity of $Q_1\subseteq Q_2$ is NP-complete [1808.08822]. This places the semantic recognition problem for MonoRel within a familiar complexity-theoretic landscape: expressive collapse does not imply trivial recognizability.

The paper also places the result in the broader tradition of homomorphism-preservation and Lyndon-type theorems in finite model theory, specialized here to CQ containments [1808.08822]. This suggests that MonoRel is part of a wider family of phenomena in which monotonicity or positivity constraints force syntactic simplification.

## 6. Terminological scope in adjacent literatures

The term **monotone relation** is used in several technically distinct research areas, and these usages should not be conflated with the database-theoretic notion of MonoRel. In Banach-space operator theory, a linear relation $A:X\rightrightarrows X^*$ is monotone when
\[
\langle x-y,\ x^*-y^*\rangle\ge 0
\]
for all graph points, and for maximally monotone linear relations the dense type, negative-infimum type, and Fitzpatrick–Phelps type are equivalent and coincide with monotonicity of the adjoint [1103.6239]. In finite-dimensional Euclidean spaces, monotone linear relations admit explicit maximally monotone linear subspace extensions characterized through the symmetric matrix $AB^T+BA^T$ [1103.1409], and the relationships among strict, paramonotone, $3$-cyclic, $3^*$, and maximal monotonicity have been analyzed systematically for linear relations in Hilbert space [1211.1048]. In abstract PDE theory, elliptic boundary-value problems in divergence form can be reformulated as inversion of a maximal monotone relation in a Hilbert space [1209.0402], while stochastic homogenization of random maximal monotone operators uses Fitzpatrick’s variational formulation, Visintin’s scale integration/disintegration theory, and Tartar–Murat compensated compactness [1702.08532].

In metric geometry, monotone relations have also been defined for Hadamard spaces. There, a relation $M\subset X\times X^\lozenge$ is monotone if
\[
\langle x^\lozenge-y^\lozenge,\overrightarrow{yx}\rangle\ge 0
\]
for all $(x,x^\lozenge),(y,y^\lozenge)\in M$, and with the $\mathcal W$-property one obtains the characterization
\[
M\text{ is monotone}\iff \Theta_M=\varsigma_M
\]
[1906.00396].

In model theory and finite-variable logic, “monotone” binary relations on linear orders are interval-preserving relations. Over linear orders expanded by such relations, monadic first-order logic has the three-variable property, yielding $\mathrm{FO}=\mathrm{FO}^3$ [1904.00189].

These parallel usages underscore that MonoRel in database theory is a domain-specific abbreviation: it names the monotone fragment of containment-of-CQ Boolean queries, not the general theory of monotone multifunctions or interval-preserving binary relations.

Source: https://www.emergentmind.com/topics/monotone-relations-monorel