Normal Modal Logics

Filtrations and Decidability

Reading preferences

Optional display controls need JavaScript. All reading content and navigation work without it.

Source file content/normal-modal-logic/filtrations/filtrations.tex

Source file content/normal-modal-logic/filtrations/introduction.tex

Introduction

One important question about a logic is always whether it is decidable, i.e., if there is an effective procedure which will answer the question “is this formula valid.” Propositional logic is decidable: we can effectively test if a formula is a tautology by constructing a truth table, and for a given formula, the truth table is finite. But we can't obviously test if a modal formula is true in all models, for there are infinitely many of them. We can list all the finite models relevant to a given formula, since only the assignment of subsets of worlds to propositional variables which actually occur in the formula are relevant. If the accessibility relation is fixed, the possible different assignments V(p)V(p)source are just all the subsets of WWsource, and if |W|=n\card{W} = nsource there are 2n2^nsource of those. If our formula A!Asource contains mmsource propositional variables there are then 2nm2^{nm}source different models with nnsource worlds. For each one, we can test if A!Asource is true at all worlds, simply by computing the truth value of A!Asource in each. Of course, we also have to check all possible accessibility relations, but there are only finitely many relations on nnsource worlds as well (specifically, the number of subsets of W×WW \times Wsource, i.e., 2n22^{n^2}source.

If we are not interested in the logic LogK, but a logic defined by some class of models (e.g., the reflexive transitive models), we also have to be able to test if the accessibility relation is of the right kind. We can do that whenever the frames we are interested in are definable by modal formulas (e.g., by testing if AxT and Ax4 valid in the frame). So, the idea would be to run through all the finite frames, test each one if it is a frame in the class we're interested in, then list all the possible models on that frame and test if A!Asource is true in each. If not, stop: A!Asource is not valid in the class of models of interest.

There is a problem with this idea: we don't know when, if ever, we can stop looking. If the formula has a finite countermodel, our procedure will find it. But if it has no finite countermodel, we won't get an answer. The formula may be valid (no countermodels at all), or it may have only an infinite countermodel, which we'll never look at. This problem can be overcome if we can show that every formula that has a countermodel has a finite countermodel. If this is the case we say the logic has the finite model property.

But how would we show that a logic has the finite model property? One way of doing this would be to find a way to turn an infinite (counter)model of A!Asource into a finite one. If that can be done, then whenever there is a model in which A!Asource is not true, then the resulting finite model also makes A!Asource not true. That finite model will show up on our list of all finite models, and we will eventually determine, for every formula that is not valid, that it isn't. Our procedure won't terminate if the formula is valid. If we can show in addition that there is some maximum size that the finite model our procedure provides can have, and that this maximum size depends only on the formula A!Asource, we will have a size up to which we have to test finite models in our search for countermodels. If we haven't found a countermodel by then, there are none. Then our procedure will, in fact, decide the question “is A!Asource valid?” for any formula A!Asource.

A strategy that often works for turning infinite structures into finite structures is that of “identifying” elements of the structure which behave the same way in relevant respects. If there are infinitely many worlds in M\mModel{M}source that behave the same in relevant respects, then we may be able to collect all worlds in finitely many (possibly infinite) “classes” of such worlds. In other words, we should partition the set of worlds in the right way, i.e., in such a way that each partition contains infinitely many worlds, but there are only finitely many partitions. Then we define a new model M*\mModel{M^*}source where the worlds are the partitions. Finitely many partitions in the old model give us finitely many worlds in the new model, i.e., a finite model. Let's call the partition a world wwsource is in [w][w]source. We'll want it to be the case that MA[w]\mSat{M}{!A}[w]source iff M*A[[w]]\mSat{M^*}{!A}[{[w]}]source, since we want the new model to be a countermodel to A!Asource if the old one was. This requires that we define the partition, as well as the accessibility relation of M*\mModel{M^*}source in the right way.

To see how this would go, first imagine we have no accessibility relation. MB[w]\mSat{M}{\Box !B}[w]source iff for some vWv \in Wsource, MB[v]\mSat{M}{\Box !B}[v]source, and the same for M*\mModel{M^*}source, except with [w][w]source and [v][v]source. As a first idea, let's say that two worlds uusource and vvsource are equivalent (belong to the same partition) if they agree on all propositional variables in M\mModel{M}source, i.e., Mp[u]\mSat{M}{p}[u]source iff Mp[v]\mSat{M}{p}[v]source. Let V*(p)={[w]:Mp[w]}V^*(p) = \Setabs{[w]}{\mSat{M}{p}[w]}source. Our aim is to show that MA[w]\mSat{M}{!A}[w]source iff M*A[[w]]\mSat{M^*}{!A}[{[w]}]source. Obviously, we'd prove this by induction: The base case would be Ap!A \equiv psource. First suppose Mp[w]\mSat{M}{p}[w]source. Then [w]V*[w] \in V^*source by definition, so M*p[[w]]\mSat{M^*}{p}[{[w]}]source. Now suppose that M*p[[w]]\mSat{M^*}{p}[{[w]}]source. That means that [w]V*(p)[w] \in V^*(p)source, i.e., for some vvsource equivalent to wwsource, Mp[v]\mSat{M}{p}[v]source. But “wwsource equivalent to vvsource” means “wwsource and vvsource make all the same propositional variables true,” so Mp[w]\mSat{M}{p}[w]source. Now for the inductive step, e.g., A¬B!A \ident \lnot !Bsource. Then M¬B[w]\mSat{M}{\lnot !B}[w]source iff MB[w]\mSat/{M}{!B}[w]source iff M*B[[w]]\mSat/{M^*}{!B}[{[w]}]source (by inductive hypothesis) iff M*¬B[[w]]\mSat{M^*}{\lnot !B}[{[w]}]source. Similarly for the other non-modal operators. It also works for \Boxsource: suppose M*B[[w]]\mSat{M^*}{\Box !B}[{[w]}]source. That means that for every [u][u]source, M*B[[u]]\mSat{M^*}{!B}[{[u]}]source. By inductive hypothesis, for every uusource, MB[u]\mSat{M}{!B}[u]source. Consequently, MB[w]\mSat{M}{\Box !B}[w]source.

In the general case, where we have to also define the accessibility relation for M*\mModel{M^*}source, things are more complicated. We'll call a model M*\mModel{M^*}source a filtration if its accessibility relation R*R^*source satisfies the conditions required to make the inductive proof above go through. Then any filtration M*\mModel{M^*}source will make A!Asource true at [w][w]source iff M\mModel{M}source makes A!Asource true at wwsource. However, now we also have to show that there are filtrations, i.e., we can define R*R^*source so that it satisfies the required conditions. In order for this to work, however, we have to require that worlds uusource, vvsource count as equivalent not just when they agree on all propositional variables, but on all sub-formulas of A!Asource. Since A!Asource has only finitely many sub-formulas, this will still guarantee that the filtration is finite. There is not just one way to define a filtration, and in order to make sure that the accessibility relation of the filtration satisfies the required properties (e.g., reflexive, transitive, etc.) we have to be inventive with the definition of R*R^*source.

Source file content/normal-modal-logic/filtrations/preliminaries.tex

Preliminaries

Filtrations allow us to establish the decidability of our systems of modal logic by showing that they have the finite model property, i.e., that any formula that is true (false) in a model is also true (false) in a finite model. Filtrations are defined relative to sets of formulas which are closed under subformulas.

Definition of subformula and modal closure

A set Γ\Gammasource of formulas is closed under subformulas if it contains every subformula of a formula in Γ\Gammasource. Further, Γ\Gammasource is modally closed if it is closed under subformulas and moreover AΓ!A \in \Gammasource implies A,AΓ\Box!A, \Diamond!A \in \Gammasource.

For instance, given a formula A!Asource, the set of all its sub-formulas is closed under sub-formulas. When we're defining a filtration of a model through the set of sub-formulas of A!Asource, it will have the property we're after: it makes A!Asource true (false) iff the original model does.

The set of worlds of a filtration of M\mModel{M}source through Γ\Gammasource is defined as the set of all equivalence classes of the following equivalence relation.

Filtration equivalence relation

Let M=W,R,V\mModel{M} =\tuple{W, R, V}source and suppose Γ\Gammasource is closed under sub-formulas. Define a relation \equivsource on WWsource to hold of any two worlds that make the same formulas from Γ\Gammasource true, i.e.:

uvif and only if AΓ:MA[u]MA[v].u \equiv v \quad \text{if and only if } \quad \forall !A \in \Gamma : \mSat{M}{!A}[u] \Leftrightarrow \mSat{M}{!A}[v].source

The equivalence class [w][w]_\equivsource of a world wwsource, or [w][w]source for short, is the set of all worlds \equivsource-equivalent to wwsource:

[w]={v:vw}.[w] = \Setabs{v}{v \equiv w}.source

The filtration relation is an equivalence relation

Given M\mModel{M}source and Γ\Gammasource, \equivsource as defined above is an equivalence relation, i.e., it is reflexive, symmetric, and transitive.

Proof

The relation \equivsource is reflexive, since wwsource makes exactly the same formulas from Γ\Gammasource true as itself. It is symmetric since if uusource makes the same formulas from Γ\Gammasource true as vvsource, the same holds for vvsource and uusource. It is also transitive, since if uusource makes the same formulas from Γ\Gammasource true as vvsource, and vvsource as wwsource, then uusource makes the same formulas from Γ\Gammasource true as wwsource.

The relation \equivsource, like any equivalence relation, divides WWsource into partitions, i.e., subsets of WWsource which are pairwise disjoint, and together cover all of WWsource. Every wWw \in Wsource is an element of one of the partitions, namely of [w][w]source, since www \equiv wsource. So the partitions [w][w]source cover all of WWsource. They are pairwise disjoint, for if u[w]u \in [w]source and u[v]u \in [v]source, then uwu \equiv wsource and uvu \equiv vsource, and by symmetry and transitivity, wvw \equiv vsource, and so [w]=[v][w] = [v]source.

Source file content/normal-modal-logic/filtrations/filtrations-def.tex

Filtrations

Rather than define “the” filtration of M\mModel{M}source through Γ\Gammasource, we define when a model M*\mModel{M^*}source counts as a filtration of M\mModel{M}source. All filtrations have the same set of worlds W*W^*source and the same valuation V*V^*source. But different filtrations may have different accessibility relations R*R^*source. To count as a filtration, R*R^*source has to satisfy a number of conditions, however. These conditions are exactly what we'll require to prove the main result, namely that MA[w]\mSat{M}{!A}[w]source iff M*A[[w]]\mSat{M^*}{!A}[{[w]}]source, provided AΓ!A \in \Gammasource.

Definition of a filtration

Let Γ\Gammasource be closed under subformulas and M=W,R,V\mModel{M} = \tuple{W, R, V}source. A filtration of M\mModel{M}source through Γ\Gammasource is any model M*=W*,R*,V*\mModel{M^*} = \tuple{W^*,R^*,V^*}source, where:

  1. W*={[w]:wW}W^* = \Setabs{[w]}{w \in W}source;

  2. For any u,vWu,v \in Wsource:

    1. If RuvRuvsource then R*[u][v]R^*[u][v]source;

    2. If R*[u][v]R^*[u][v]source then for any AΓ\Box !A \in \Gammasource, if MA[u]\mSat{M}{\Box!A}[u]source then MA[v]\mSat{M}{!A}[v]source;

    3. If R*[u][v]R^*[u][v]source then for any AΓ\Diamond !A \in \Gammasource, if MA[v]\mSat{M}{!A}[v]source then MA[u]\mSat{M}{\Diamond!A}[u]source.

  3. V*(p)={[u]:uV(p)}V^*(p) = \Setabs{[u]}{u \in V(p)}source.

It's worthwhile thinking about what V*(p)V^*(p)source is: the set consisting of the equivalence classes [w][w]source of all worlds wwsource where ppsource is true in M\mModel{M}source. On the one hand, if wV(p)w \in V(p)source, then [w]V*(p)[w] \in V^*(p)source by that definition. However, it is not necessarily the case that if [w]V*(p)[w] \in V^*(p)source, then wV(p)w \in V(p)source. If [w]V*(p)[w] \in V^*(p)source we are only guaranteed that [w]=[u][w] = [u]source for some uV(p)u \in V(p)source. Of course, [w]=[u][w] = [u]source means that wuw \equiv usource. So, when [w]V*(p)[w] \in V^*(p)source we can (only) conclude that wuw \equiv usource for some uV(p)u \in V(p)source.

Filtration Theorem

If M*\mModel{M^*}source is a filtration of M\mModel{M}source through Γ\Gammasource, then for every AΓ!A \in \Gammasource and wWw \in Wsource, we have MA[w]\mSat{M}{!A}[w]source if and only if M*A[[w]]\mSat{M^*}{!A}[{[w]}]source.

Proof

By induction on A!Asource, using the fact that Γ\Gammasource is closed under subformulas. Since AΓ!A \in \Gammasource and Γ\Gammasource is closed under sub-formulas, all sub-formulas of A!Asource are also Γ\in \Gammasource. Hence in each inductive step, the induction hypothesis applies to the sub-formulas of A!Asource.

  1. Case: A!A \ident \lfalsesource

    Neither MA[w]\mSat{M}{\indfrm}[w]source nor M*A[[w]]\mSat{M^*}{\indfrm}[{[w]}]source.

  2. Case: Ap!A \ident psource

    The left-to-right direction is immediate, as MA[w]\mSat{M}{\indfrm}[w]source only if wV(p)w \in V(p)source, which implies [w]V*(p)[w] \in V^*(p)source, i.e., M*A[[w]]\mSat{M^*}{\indfrm}[{[w]}]source. Conversely, suppose M*A[[w]]\mSat{M^*}{\indfrm}[{[w]}]source, i.e., [w]V*(p)[w] \in V^*(p)source. Then for some vV(p)v \in V(p)source, wvw \equiv vsource. Of course then also Mp[v]\mSat{M}{p}[v]source. Since wvw \equiv vsource, wwsource and vvsource make the same formulas from Γ\Gammasource true. Since by assumption pΓp \in \Gammasource and Mp[v]\mSat{M}{p}[v]source, MA[w]\mSat{M}{\indfrm}[w]source.

  3. Case: A¬B!A \ident \lnot !Bsource

    MA[w]\mSat{M}{\indfrm}[w]source iff MB[w]\mSat/{M}{!B}[w]source. By induction hypothesis, MB[w]\mSat/{M}{!B}[w]source iff M*B[[w]]\mSat/{M^*}{!B}[{[w]}]source. Finally, M*B[[w]]\mSat/{M^*}{!B}[{[w]}]source iff M*A[[w]]\mSat{M^*}{\indfrm}[{[w]}]source.

  4. Exercise.

  5. Case: A(BC)!A \ident (!B \lor !C)source

    MA[w]\mSat{M}{\indfrm}[w]source iff MB[w]\mSat{M}{!B}[w]source or MC[w]\mSat{M}{!C}[w]source. By induction hypothesis, MB[w]\mSat{M}{!B}[w]source iff M*B[[w]]\mSat{M^*}{!B}[{[w]}]source, and MC[w]\mSat{M}{!C}[w]source iff M*C[[w]]\mSat{M^*}{!C}[{[w]}]source. And M*A[[w]]\mSat{M^*}{\indfrm}[{[w]}]source iff M*B[[w]]\mSat{M^*}{!B}[{[w]}]source or M*C[[w]]\mSat{M^*}{!C}[{[w]}]source.

  6. Exercise.

  7. Case: AB!A \ident \Box !Bsource

    Suppose MA[w]\mSat{M}{\indfrm}[w]source; to show that M*A[[w]]\mSat{M^*}{\indfrm}[{[w]}]source, let vvsource be such that R*[w][v]R^*[w][v]source. From the definition of a filtrationthe necessity-preservation condition for a filtration, we have that MB[v]\mSat{M}{!B}[v]source, and by inductive hypothesis M*B[[v]]\mSat{M^*}{!B}[{[v]}]source. Since vvsource was arbitrary, M*A[[w]]\mSat{M^*}{\indfrm}[{[w]}]source follows.

    Conversely, suppose M*A[[w]]\mSat{M^*}{\indfrm}[{[w]}]source and let vvsource be arbitrary such that RwvRwvsource. From the definition of a filtrationthe inherited-edge condition for a filtration, we have R*[w][v]R^*[w][v]source, so that M*B[[v]]\mSat{M^*}{!B}[{[v]}]source; by inductive hypothesis MB[v]\mSat{M}{!B}[v]source, and since vvsource was arbitrary, MA[w]\mSat{M}{\indfrm}[w]source.

  8. Exercise.

Exercise completing the Filtration Theorem

Complete the proof of the Filtration Theorem preserving truth of formulas in Gamma

What holds for truth at worlds in a model also holds for truth in a model and validity in a class of models.

Model truth and class validity under filtration

Let Γ\Gammasource be closed under subformulas. Then:

  1. If M*\mModel{M^*}source is a filtration of M\mModel{M}source through Γ\Gammasource then for any AΓ!A \in \Gammasource: MA\mSat{M}{!A}source if and only if M*A\mSat{M^*}{!A}source.

  2. If C\mClass{C}source is a class of models and Γ(C)\Gamma(\mClass{C})source is the class of Γ\Gammasource-filtrations of models in C\mClass{C}source, then any formula AΓ!A \in \Gammasource is valid in C\mClass{C}source if and only if it is valid in Γ(C)\Gamma(\mClass{C})source.

Source file content/normal-modal-logic/filtrations/examples-of-filtrations.tex

Examples of Filtrations

We have not yet shown that there are any filtrations. But indeed, for any model M\mModel{M}source, there are many filtrations of M\mModel{M}source through Γ\Gammasource. We identify two, in particular: the finest and coarsest filtrations. Filtrations of the same models will differ in their accessibility relation (as the definition of a filtration stipulates directly what W*W^*source and V*V^*source should be). The finest filtration will have as few related worlds as possible, whereas the coarsest will have as many as possible.

Definition of the finest filtration

Where Γ\Gammasource is closed under subformulas, the finest filtration M*\mModel{M^*}source of a model M\mModel{M}source is defined by putting:

R*[u][v]if and only ifu[u]v[v]:Ruv.R^*[u][v] \quad \text{if and only if} \quad \exists u'\in [u] \; \exists v' \in [v] : Ru'v'.source

The finest construction is a filtration

The finest filtration M*\mModel{M^*}source is indeed a filtration.

Proof

We need to check that R*R^*source, so defined, satisfies the definition of a filtrationthe accessibility conditions in the definition of a filtration. We check the three conditions in turn.

If RuvRuvsource then since u[u]u \in [u]source and v[v]v \in [v]source, also R*[u][v]R^*[u][v]source, so the inherited-edge condition for a filtration is satisfied.

For the necessity-preservation condition for a filtration, suppose AΓ\Box!A \in \Gammasource, R*[u][v]R^*[u][v]source, and MA[u]\mSat{M}{\Box!A}[u]source. By definition of R*R^*source, there are uuu' \equiv usource and vvv' \equiv vsource such that RuvRu'v'source. Since uusource and uu'source agree on Γ\Gammasource, also MA[u]\mSat{M}{\Box!A}[u']source, so that MA[v]\mSat{M}{!A}[v']source. By closure of Γ\Gammasource under sub-formulas, vvsource and vv'source agree on A!Asource, so MA[v]\mSat{M}{!A}[v]source, as desired.

We leave the verification of the possibility-preservation condition for a filtration as an exercise.

Exercise completing the finest-filtration proof

probBox,probDiamond Complete the proof of the proposition that the finest construction is a filtration.

Definition of the coarsest filtration

Where Γ\Gammasource is closed under subformulas, the coarsest filtration M*\mModel{M^*}source of a model M\mModel{M}source is defined by putting R*[u][v]R^*[u][v]source if and only if

both of the following conditions are met:

Tagenumerate

prvBox,prvDiamond item If AΓ\Box!A \in \Gammasource and MA[u]\mSat{M}{\Box!A}[u]source then MA[v]\mSat{M}{!A}[v]source ; item If AΓ\Diamond!A \in \Gammasource and MA[v]\mSat{M}{!A}[v]source then MA[u]\mSat{M}{\Diamond!A}[u]source.

The coarsest construction is a filtration

The coarsest filtration M*\mModel{M^*}source is indeed a filtration.

Proof

Given the definition of R*R^*source, the only condition that is left to verify is the implication from RuvRuvsource to R*[u][v]R^*[u][v]source. So assume RuvRuvsource. Suppose AΓ\Box!A \in \Gammasource and MA[u]\mSat{M}{\Box!A}[u]source; then obviously MA[v]\mSat{M}{!A}[v]source , and the box clause in the definition of the coarsest filtration is satisfied. Suppose AΓ\Diamond!A \in \Gammasource and MA[v]\mSat{M}{!A}[v]source. Then MA[u]\mSat{M}{\Diamond!A}[u]source since RuvRuvsource , and the diamond clause in the definition of the coarsest filtration is satisfied .

Infinite alternating model and its filtrations

Let W=Z+W = \PosIntsource, RnmRnmsource iff m=n+1m = n + 1source, and V(p)={2n:n}V(p) = \Setabs{2n}{n \in \Nat}source. The model M=W,R,V\mModel{M} = \tuple{W, R, V}source is depicted in the figure showing the infinite alternating model and its two filtrations. The worlds are 11source, 22source, etc.; each world can access exactly one other world---its successor---and ppsource is true at all and only the even numbers.

Figure of the alternating model and two filtrations

The outer figure contains one infinite-chain diagram and one diagram showing the finest and coarsest two-class filtrations. The two inner diagrams carry the complete node, valuation, edge, loop, and omission descriptions.

Source transcription

centering

Infinite alternating chain model

Infinite alternating successor-chain model. Nodes in source order. Source world node world one. Valuation label p is false. Printed node label one. Source world node world two. Valuation label p is true. Printed node label two. Source world node world three. Valuation label p is false. Printed node label three. Source world node world four. Valuation label p is true. Printed node label four. Source phantom node phantom marker five is an unlabeled layout marker and is not an actual world. Directed edges in source order. Directed accessibility edge from world one to world two. Directed accessibility edge from world two to world three. Directed accessibility edge from world three to world four. Omissions and continuation marks. A dotted, non-arrow line from world four to phantom marker five indicates that the successor chain continues; it is not recorded as a directed accessibility edge to a named world. Only the three solid directed arrows are accessibility edges. The dotted line has no arrowhead and records continuation without naming another actual world. End diagram.

Nodes

  1. Node 1: one¬p\mFalse{p}source11source
  2. Node 2: twop\mTrue{p}source22source
  3. Node 3: three¬p\mFalse{p}source33source
  4. Node 4: fourp\mTrue{p}source44source
  5. Node 5: unlabeled phantom continuation marker

Edges

  1. Edge 1: 1 to 2; accessibility relation R.
  2. Edge 2: 2 to 3; accessibility relation R.
  3. Edge 3: 3 to 4; accessibility relation R.
source 110

Two two-class filtrations

Two separate two-class filtrations printed in one TikZ picture. Nodes in source order. In the first two-class filtration, Source world node first-model odd-class node. Valuation label p is false. Printed node label the equivalence class of one. In the first two-class filtration, Source world node first-model even-class node. Valuation label p is true. Printed node label the equivalence class of two. In the second two-class filtration, Source world node second-model odd-class node. Valuation label p is false. Printed node label the equivalence class of one. In the second two-class filtration, Source world node second-model even-class node. Valuation label p is true. Printed node label the equivalence class of two. Directed edges in source order. Directed accessibility edge from first-model odd-class node to first-model even-class node. Directed accessibility edge from first-model even-class node to first-model odd-class node. Directed accessibility edge from second-model odd-class node to second-model even-class node. Directed accessibility edge from second-model even-class node to second-model odd-class node. Directed accessibility loop at second-model even-class node. Omissions and continuation marks. There are no edges between the first and second displayed components. The second component alone has the printed loop at its even class; no other loop is inferred. End diagram.

Nodes

  1. Node 1: the equivalence class of one¬p\mFalse{p}source[1][1]source
  2. Node 2: the equivalence class of twop\mTrue{p}source[2][2]source
  3. Node 3: the equivalence class of one¬p\mFalse{p}source[1][1]source
  4. Node 4: the equivalence class of twop\mTrue{p}source[2][2]source

Edges

  1. Edge 1: 1 to 2; accessibility relation R star.
  2. Edge 2: 2 to 1; accessibility relation R star.
  3. Edge 3: 3 to 4; accessibility relation R star.
  4. Edge 4: 4 to 3; accessibility relation R star.
  5. Edge 5: 4 to 4; accessibility relation R star; loop.
source 122

captionAn infinite model and its filtrations.

Now let Γ\Gammasource be the set of sub-formulas of pp\Box p \lif psource, i.e., {p,p,pp}\{p, \Box p, \Box p \lif p\}source. ppsource is true at all and only the even numbers, p\Box psource is true at all and only the odd numbers, so pp\Box p \lif psource is true at all and only the even numbers. In other words, every odd number makes p\Box psource true but ppsource and pp\Box p \lif psource false; every even number makes ppsource and pp\Box p \lif psource true, but p\Box psource false. So W*={[1],[2]}W^* = \{ [1], [2] \}source, where [1]={1,3,5,}[1] = \{1, 3, 5, \dots\}source and [2]={2,4,6,}[2] = \{2, 4, 6, \dots\}source. Since 2V(p)2 \in V(p)source, [2]V*(p)[2] \in V^*(p)source; since 1V(p)1 \notin V(p)source, [1]V*(p)[1] \notin V^*(p)source. So V*(p)={[2]}V^*(p) = \{[2]\}source.

Any filtration based on W*W^*source must have an accessibility relation that includes [1],[2],[2],[1]\tuple{[1], [2]}, \tuple{[2],[1]}source: since R12R12source, we must have R*[1][2]R^*[1][2]source by the definition of a filtrationthe inherited-edge condition for a filtration, and since R23R23source we must have R*[2][3]R^*[2][3]source, and [3]=[1][3]=[1]source. It cannot include [1],[1]\tuple{[1],[1]}source: if it did, we'd have R*[1][1]R^*[1][1]source, Mp[1]\mSat{M}{\Box p}[1]source but Mp[1]\mSat/{M}{p}[1]source, contradicting the necessity-preservation condition for a filtration. Nothing requires or rules out that R*[2][2]R^*[2][2]source. So, there are two possible filtrations of M\mModel{M}source, corresponding to the two accessibility relations

{[1],[2],[2],[1]} and {[1],[2],[2],[1],[2],[2]}.\{\tuple{[1],[2]}, \tuple{[2],[1]}\} \text{ and } \{\tuple{[1],[2]}, \tuple{[2],[1]}, \tuple{[2],[2]}\}.source

In either case, ppsource and pp\Box p \lif psource are false and p\Box psource is true at [1][1]source; ppsource and pp\Box p \lif psource are true and p\Box psource is false at [2][2]source.

Exercise filtering an infinite binary tree

Consider the following model M=W,R,V\mModel{M} = \tuple{W, R, V}source where W={0σ:σB*}W = \Setabs{0\sigma}{\sigma \in \Bin^*}source, the set of sequences of 00sources and 11sources starting with 00source, with RσσR\sigma\sigma'source iff σ=σ0\sigma' = \sigma 0source or σ=σ1\sigma' = \sigma 1source, and V(p)={σ0:σB*}V(p) = \Setabs{\sigma 0}{\sigma \in \Bin^*}source and V(q)={σ1:σB*{1}}V(q) = \Setabs{\sigma 1}{\sigma \in \Bin^* \setminus \{1\}}source. Here's a picture:

Infinite binary tree model

Displayed finite prefix of an infinite binary-tree model. Nodes in source order. Source world node world zero. Valuation label p is true. Valuation label q is false. Printed node label zero. Source world node world zero zero. Valuation label p is true. Valuation label q is false. Printed node label zero zero. Source world node world zero zero zero. Valuation label p is true. Valuation label q is false. Printed node label zero zero zero. Source phantom node phantom marker zero zero zero zero is an unlabeled layout marker and is not an actual world. Source phantom node phantom marker zero zero zero one is an unlabeled layout marker and is not an actual world. Source world node world zero zero one. Valuation label p is false. Valuation label q is true. Printed node label zero zero one. Source phantom node phantom marker zero zero one zero is an unlabeled layout marker and is not an actual world. Source phantom node phantom marker zero zero one one is an unlabeled layout marker and is not an actual world. Source world node world zero one. Valuation label p is false. Valuation label q is true. Printed node label zero one. Source world node world zero one zero. Valuation label p is true. Valuation label q is false. Printed node label zero one zero. Source phantom node phantom marker zero one zero zero is an unlabeled layout marker and is not an actual world. Source phantom node phantom marker zero one zero one is an unlabeled layout marker and is not an actual world. Source world node world zero one one. Valuation label p is false. Valuation label q is true. Printed node label zero one one. Source phantom node phantom marker zero one one zero is an unlabeled layout marker and is not an actual world. Source phantom node phantom marker zero one one one is an unlabeled layout marker and is not an actual world. Directed edges in source order. Directed accessibility edge from world zero to world zero zero. Directed accessibility edge from world zero to world zero one. Directed accessibility edge from world zero zero to world zero zero zero. Directed accessibility edge from world zero zero to world zero zero one. Directed accessibility edge from world zero one to world zero one zero. Directed accessibility edge from world zero one to world zero one one. Omissions and continuation marks. A dotted, non-arrow child stub from world zero zero zero to phantom marker zero zero zero zero indicates omitted continuation of the binary tree; it is not recorded as a directed accessibility edge. A dotted, non-arrow child stub from world zero zero zero to phantom marker zero zero zero one indicates omitted continuation of the binary tree; it is not recorded as a directed accessibility edge. A dotted, non-arrow child stub from world zero zero one to phantom marker zero zero one zero indicates omitted continuation of the binary tree; it is not recorded as a directed accessibility edge. A dotted, non-arrow child stub from world zero zero one to phantom marker zero zero one one indicates omitted continuation of the binary tree; it is not recorded as a directed accessibility edge. A dotted, non-arrow child stub from world zero one zero to phantom marker zero one zero zero indicates omitted continuation of the binary tree; it is not recorded as a directed accessibility edge. A dotted, non-arrow child stub from world zero one zero to phantom marker zero one zero one indicates omitted continuation of the binary tree; it is not recorded as a directed accessibility edge. A dotted, non-arrow child stub from world zero one one to phantom marker zero one one zero indicates omitted continuation of the binary tree; it is not recorded as a directed accessibility edge. A dotted, non-arrow child stub from world zero one one to phantom marker zero one one one indicates omitted continuation of the binary tree; it is not recorded as a directed accessibility edge. Only the six solid directed arrows are recorded as accessibility edges. Each dotted child stub is an explicit continuation marker, not a solid directed edge to an actual named world. End diagram.

Nodes

  1. Node 1: zerop\mTrue{p}source¬q\mFalse{q}source00source
  2. Node 2: zero zerop\mTrue{p}source¬q\mFalse{q}source0000source
  3. Node 3: zero zero zerop\mTrue{p}source¬q\mFalse{q}source000000source
  4. Node 4: unlabeled phantom continuation marker
  5. Node 5: unlabeled phantom continuation marker
  6. Node 6: zero zero one¬p\mFalse{p}sourceq\mTrue{q}source001001source
  7. Node 7: unlabeled phantom continuation marker
  8. Node 8: unlabeled phantom continuation marker
  9. Node 9: zero one¬p\mFalse{p}sourceq\mTrue{q}source0101source
  10. Node 10: zero one zerop\mTrue{p}source¬q\mFalse{q}source010010source
  11. Node 11: unlabeled phantom continuation marker
  12. Node 12: unlabeled phantom continuation marker
  13. Node 13: zero one one¬p\mFalse{p}sourceq\mTrue{q}source011011source
  14. Node 14: unlabeled phantom continuation marker
  15. Node 15: unlabeled phantom continuation marker

Edges

  1. Edge 1: 0 to 00; accessibility relation R.
  2. Edge 2: 0 to 01; accessibility relation R.
  3. Edge 3: 00 to 000; accessibility relation R.
  4. Edge 4: 00 to 001; accessibility relation R.
  5. Edge 5: 01 to 010; accessibility relation R.
  6. Edge 6: 01 to 011; accessibility relation R.
source 177

We have M(pq)(pq)[w]\mSat/{M}{\Box(p \lor q) \lif (\Box p \lor \Box q)}[w]source for every wwsource.

Let Γ\Gammasource be the set of sub-formulas of (pq)(pq)\Box(p \lor q) \lif (\Box p \lor \Box q)source. What are W*W^*source and V*V^*source? What is the accessibility relation of the finest filtration of M\mModel{M}source? Of the coarsest?

Source file content/normal-modal-logic/filtrations/finite.tex

Filtrations are Finite

We've defined filtrations for any set Γ\Gammasource that is closed under sub-formulas. Nothing in the definition itself guarantees that filtrations are finite. In fact, when Γ\Gammasource is infinite (e.g., is the set of all formulas), it may well be infinite. However, if Γ\Gammasource is finite (e.g., when it is the set of sub-formulas of a given formula A!Asource), so is any filtration through Γ\Gammasource.

Finite size of a filtration

If Γ\Gammasource is finite then any filtration M*\mModel{M^*}source of a model M\mModel{M}source through Γ\Gammasource is also finite.

Proof

The size of W*W^*source is the number of different classes [w][w]source under the equivalence relation \equivsource. Any two worlds uusource, vvsource in such class---that is, any uusource and vvsource such that uvu \equiv vsource---agree on all formulas A!Asource in Γ\Gammasource, AΓ!A \in \Gammasource either A!Asource is true at both uusource and vvsource, or at neither. So each class [w][w]source corresponds to subset of Γ\Gammasource, namely the set of all AΓ!A \in \Gammasource such that A!Asource is true at the worlds in [w][w]source. No two different classes [u][u]source and [v][v]source correspond to the same subset of Γ\Gammasource. For if the set of formulas true at uusource and that of formulas true at vvsource are the same, then uusource and vvsource agree on all formulas in Γ\Gammasource, i.e., uvu \equiv vsource. But then [u]=[v][u] = [v]source. So, there is an injective function from W*W^*source to (Γ)\Pow{\Gamma}source, and hence |W*||(Γ)|\card{W^*} \le \card{\Pow{\Gamma}}source. Hence if Γ\Gammasource contains nnsource sentences, the cardinality of W*W^*source is no greater than 2n2^nsource.

Source file content/normal-modal-logic/filtrations/S5-fmp.tex

LogK and LogS5 have the Finite Model Property

Definition of the finite model property

A system Σ\Sigmasource of modal logic is said to have the finite model property if whenever a formula A!Asource is true at a world in a model of Σ\Sigmasource then A!Asource is true at a world in a finite model of Σ\Sigmasource.

Finite model property for K

LogK has the finite model property.

Proof

LogK is the set of valid formulas, i.e., any model is a model of LogK. By the Filtration Theorem preserving truth of formulas in Gamma, if MA[w]\mSat{M}{!A}[w]source, then M*A[w]\mSat{M^*}{!A}[w]source for any filtration of M\mModel{M}source through the set Γ\Gammasource of sub-formulas of A!Asource. Any formula only has finitely many sub-formulas, so Γ\Gammasource is finite. By the proposition bounding the size of a filtration, |W*|2n\card{W^*} \le 2^nsource, where nnsource is the number of formulas in Γ\Gammasource. And since LogK imposes no restriction on models, M*\mModel{M^*}source is a LogK-model.

To show that a logic LogL has the finite model property via filtrations it is essential that the filtration of an LogL-model is itself a LogL-model. Often this requires a fair bit of work, and not any filtration yields a LogL-model. However, for universal models, this still holds.

Universal validity reduced to finite universal models

Let U\mClass{U}source be the class of universal models (see the equivalence between S five models and universal models used here) and UFin\mClass{U}_\mathrm{Fin}source the class of all finite universal models. Then any formula A!Asource is valid in U\mClass{U}source if and only if it is valid in UFin\mClass{U}_\mathrm{Fin}source.

Proof

Finite universal models are universal models, so the left-to-right direction is trivial. For the right-to left direction, suppose that A!Asource is false at some world wwsource in a universal model M\mModel{M}source. Let Γ\Gammasource contain A!Asource as well as all of its subformulas; clearly Γ\Gammasource is finite. Take a filtration M*\mModel{M^*}source of M\mModel{M}source; then M*\mModel{M^*}source is finite by the proposition bounding the size of a filtration, and by the Filtration Theorem preserving truth of formulas in Gamma, A!Asource is false at [w][w]source in M*\mModel{M^*}source. It remains to observe that M*\mModel{M^*}source is also universal: given uusource and vvsource, by hypothesis RuvRuvsource and by the definition of a filtrationthe accessibility conditions in the definition of a filtration, also R*[u][v]R^*[u][v]source.

Finite model property for S five

LogS5 has the finite model property.

Proof

By the equivalence between S five models and universal models used here, if A!Asource is true at a world in some reflexive and euclidean model then it is true at a world in a universal model. By the proposition reducing universal-model validity to finite universal models, it is true at a world in a finite universal model (namely the filtration of the model through the set of sub-formulas of A!Asource). Every universal model is also reflexive and euclidean; so A!Asource is true at a world in a finite reflexive euclidean model.

Exercise on serial and reflexive filtrations

Show that any filtration of a serial or reflexive model is also serial or reflexive (respectively).

Exercise finding property-losing filtrations

Find a non-symmetric (non-transitive, non-euclidean) filtration of a symmetric (transitive, euclidean) model.

Source file content/normal-modal-logic/filtrations/S5-decidable.tex

LogS5 is Decidable

The finite model property gives us an easy way to show that systems of modal logic given by schemas are decidable (i.e., that there is a computable procedure to determine whether a formula is derivable in the system or not).

Decidability of S five

LogS5 is decidable.

Proof

Let A!Asource be given, and suppose the propositional variables occurring in A!Asource are among p1p_1source, dots, pkp_ksource. Since for each nnsource there are only finitely many models with nnsource worlds assigning a value to p1p_1source, dots, pkp_ksource, we can enumerate, in parallel, all the theorems of LogS5 by generating proofs in some systematic way; and all the models containing 11source, 22source, dots worlds and checking whether A!Asource fails at a world in some such model. Eventually one of the two parallel processes will give an answer, as by the general canonical-model determination theorem and the finite model property corollary for S five, either A!Asource is derivable or it fails in a finite universal model.

The above proof works for LogS5 because filtrations of universal models are automatically universal. The same holds for reflexivity and seriality, but more work is needed for other properties.

Source file content/normal-modal-logic/filtrations/more-filtrations.tex

Filtrations and Properties of Accessibility

As noted, filtrations of universal, serial, and reflexive models are always also universal, serial, or reflexive. But not every filtration of a symmetric or transitive model is symmetric or transitive, respectively. In some cases, however, it is possible to define filtrations so that this does hold. In order to do so, we proceed as in the definition of the coarsest filtration, but add additional conditions to the definition of R*R^*source. Let Γ\Gammasource be closed under sub-formulas. Consider the relations Ci(u,v)C_i(u,v)source in the table of conditions C one through C four between worlds uusource, vvsource in a model M=W,R,V\mModel{M} =\tuple{W, R, V}source. We can define R*[u][v]R^*[u][v]source on the basis of combinations of these conditions. For instance, if we stipulate that R*[u][v]R^*[u][v]source iff the condition C1(u,v)C_1(u,v)source holds, we get exactly the coarsest filtration. If we stipulate R*[u][v]R^*[u][v]source iff both C1(u,v)C_1(u,v)source and C2(u,v)C_2(u, v)source hold, we get a different filtration. It is “finer” than the coarsest since fewer pairs of worlds satisfy C1(u,v)C_1(u,v)source and C2(u,v)C_2(u,v)source than C1(u,v)C_1(u,v)source alone.

Conditions for accessibility-preserving filtrations

[ht] centering

Rows C one through C four

Table of four conditions on possible worlds for defining filtrations. The source has no printed header row; the accessible column roles are condition label and condition clause. Source row one, label condition C sub one of world u and world v, box clause: if necessarily formula A belongs to Gamma and necessarily formula A is true at world u in model M, then formula A is true at world v in model M. Source row two, condition C one continued, diamond clause: if possibly formula A belongs to Gamma and formula A is true at world v in model M, then possibly formula A is true at world u in model M. Source row three, label condition C sub two of world u and world v, box clause: if necessarily formula A belongs to Gamma and necessarily formula A is true at world v in model M, then formula A is true at world u in model M. Source row four, condition C two continued, diamond clause: if possibly formula A belongs to Gamma and formula A is true at world u in model M, then possibly formula A is true at world v in model M. Source row five, label condition C sub three of world u and world v, box clause: if necessarily formula A belongs to Gamma and necessarily formula A is true at world u in model M, then necessarily formula A is true at world v in model M. Source row six, condition C three continued, diamond clause: if possibly formula A belongs to Gamma and possibly formula A is true at world v in model M, then possibly formula A is true at world u in model M. Source row seven, label condition C sub four of world u and world v, box clause: if necessarily formula A belongs to Gamma and necessarily formula A is true at world v in model M, then necessarily formula A is true at world u in model M. Source row eight, condition C four continued, diamond clause: if possibly formula A belongs to Gamma and possibly formula A is true at world u in model M, then possibly formula A is true at world v in model M. End table.

Rows C one through C four
condition labelbox or diamond condition clause
C1(u,v)C_1(u,v)sourceAΓ\Box!A \in \GammasourceMA[u]\mSat{M}{\Box!A}[u]sourceMA[v]\mSat{M}{!A}[v]source
condition C one continued from the multirow labelAΓ\Diamond!A \in \GammasourceMA[v]\mSat{M}{!A}[v]sourceMA[u]\mSat{M}{\Diamond!A}[u]source
C2(u,v)C_2(u,v)sourceAΓ\Box!A \in \GammasourceMA[v]\mSat{M}{\Box!A}[v]sourceMA[u]\mSat{M}{!A}[u]source
condition C two continued from the multirow labelAΓ\Diamond!A \in \GammasourceMA[u]\mSat{M}{!A}[u]sourceMA[v]\mSat{M}{\Diamond!A}[v]source
C3(u,v)C_3(u,v)sourceAΓ\Box!A \in \GammasourceMA[u]\mSat{M}{\Box!A}[u]sourceMA[v]\mSat{M}{\Box!A}[v]source
condition C three continued from the multirow labelAΓ\Diamond!A \in \GammasourceMA[v]\mSat{M}{\Diamond!A}[v]sourceMA[u]\mSat{M}{\Diamond!A}[u]source
C4(u,v)C_4(u,v)sourceAΓ\Box!A \in \GammasourceMA[v]\mSat{M}{\Box!A}[v]sourceMA[u]\mSat{M}{\Box!A}[u]source
condition C four continued from the multirow labelAΓ\Diamond!A \in \GammasourceMA[u]\mSat{M}{\Diamond!A}[u]sourceMA[v]\mSat{M}{\Diamond!A}[v]source
source 32

captionConditions on possible worlds for defining filtrations.

Filtrations with selected accessibility properties

Let M=W,R,V\mModel{M} =\tuple{W,R,V}source be a model, Γ\Gammasource closed under sub-formulas. Let W*W^*source and V*V^*source be defined as in the definition of a filtration. Then:

  1. Suppose R*[u][v]R^*[u][v]source if and only if C1(u,v)C2(u,v)C_1(u, v) \land C_2(u,v)source. Then R*R^*source is symmetric, and M*=W*,R*,V*\mModel{M^*} = \tuple{W^*,R^*,V^*}source is a filtration if M\mModel{M}source is symmetric.

  2. Suppose R*[u][v]R^*[u][v]source if and only if C1(u,v)C3(u,v)C_1(u, v) \land C_3(u,v)source. Then R*R^*source is transitive, and M*=W*,R*,V*\mModel{M^*}=\tuple{W^*,R^*,V^*}source is a filtration if M\mModel{M}source is transitive.

  3. Suppose R*[u][v]R^*[u][v]source if and only if C1(u,v)C2(u,v)C3(u,v)C4(u,v)C_1(u, v) \land C_2(u,v) \land C_3(u,v) \land C_4(u,v)source. Then R*R^*source is symmetric and transitive, and M*=W*,R*,V*\mModel{M^*}=\tuple{W^*,R^*,V^*}source is a filtration if M\mModel{M}source is symmetric and transitive.

  4. Suppose R*R^*source is defined as R*[u][v]R^*[u][v]source if and only if C1(u,v)C3(u,v)C4(u,v)C_1(u, v) \land C_3(u,v) \land C_4(u,v)source. Then R*R^*source is transitive and euclidean, and M*=W*,R*,V*\mModel{M^*}=\tuple{W^*,R^*,V^*}source is a filtration if M\mModel{M}source is transitive and euclidean.

Proof

  1. It's immediate that R*R^*source is symmetric, since C1(u,v)C2(v,u)C_1(u,v) \Leftrightarrow C_2(v,u)source and C2(u,v)C1(v,u)C_2(u,v) \Leftrightarrow C_1(v,u)source. So it's left to show that if M\mModel{M}source is symmetric then M*\mModel{M^*}source is a filtration through Γ\Gammasource. Condition C1(u,v)C_1(u,v)source guarantees that

    the necessity-preservation condition for a filtration and the possibility-preservation condition for a filtration of the definition of a filtration are

    satisfied. So we just have to verify the definition of a filtrationthe inherited-edge condition for a filtration, i.e., that RuvRuvsource implies R*[u][v]R^*[u][v]source.

    So suppose RuvRuvsource. To show R*[u][v]R^*[u][v]source we need to establish that C1(u,v)C_1(u,v)source and C2(u,v)C_2(u,v)source. For C1C_1source: if AΓ\Box!A \in\Gammasource and MA[u]\mSat{M}{\Box!A}[u]source then also MA[v]\mSat{M}{!A}[v]source (since RuvRuvsource). Similarly, if AΓ\Diamond!A \in \Gammasource and MA[v]\mSat{M}{!A}[v]source then MA[u]\mSat{M}{\Diamond!A}[u]source since RuvRuvsource. For C2C_2source: if AΓ\Box!A \in \Gammasource and MA[v]\mSat{M}{\Box!A}[v]source then RuvRuvsource implies RvuRvusource by symmetry, so that MA[u]\mSat{M}{!A}[u]source. Similarly, if AΓ\Diamond!A \in\Gammasource and MA[u]\mSat{M}{!A}[u]source then MA[v]\mSat{M}{\Diamond!A}[v]source (since RvuRvusource by symmetry).

  2. Exercise.

  3. Exercise.

  4. Exercise.

Exercise completing the accessibility proof

Complete the proof of the theorem constructing filtrations with selected accessibility properties.

Source file content/normal-modal-logic/filtrations/euclidean-filtrations.tex

Filtrations of Euclidean Models

The approach of the section on filtrations and accessibility properties does not work in the case of models that are euclidean or serial and euclidean. Consider the model at the top of the displayed source model described as serial and euclidean, which is both euclidean and serial. Let Γ={p,p}\Gamma = \{p, \Box p \}source. When taking a filtration through Γ\Gammasource, then [w1]=[w3][w_1] = [w_3]source since w1w_1source and w3w_3source are the only worlds that agree on Γ\Gammasource. Any filtration will also have the arrow inherited from M\mModel{M}source, as depicted in the displayed filtration of that model. That model isn't euclidean. Moreover, we cannot add arrows to that model in order to make it euclidean. We would have to add double arrows between [w2][w_2]source and [w4][w_4]source, and then also between w2w_2source and w5w_5source. But p\Box psource is supposed to be true at w2w_2source, while ppsource is false at w5w_5source.

Figure called a serial and euclidean model

The outer figure contains the first five-world source graph. Its inner diagram lists all five worlds, truth labels, directed edges, and loops. The source calls the graph serial, although no outgoing edge from w sub two is drawn; no loop or edge is invented.

Source transcription

[htpb] centering

Five-world euclidean source graph

Five-world graph captioned by the source as a serial and euclidean model. Nodes in source order. Source world node world w sub one. Valuation label p is false. Modal-status label necessarily p is true at this displayed world. Printed node label world w sub one. Source world node world w sub two. Valuation label p is true. Modal-status label necessarily p is true at this displayed world. Printed node label world w sub two. Source world node world w sub three. Valuation label p is false. Modal-status label necessarily p is true at this displayed world. Printed node label world w sub three. Source world node world w sub four. Valuation label p is true. Modal-status label necessarily p is false at this displayed world. Printed node label world w sub four. Source world node world w sub five. Valuation label p is false. Modal-status label necessarily p is false at this displayed world. Printed node label world w sub five. Directed edges in source order. Directed accessibility edge from world w sub one to world w sub two. Directed accessibility edge from world w sub three to world w sub four. Directed accessibility loop at world w sub four. Directed accessibility edge from world w sub four to world w sub five. Directed accessibility edge from world w sub five to world w sub four. Directed accessibility loop at world w sub five. Omissions and continuation marks. The edge list is exactly the printed graph. In particular, no outgoing edge or loop is drawn from world w sub two, so none is added despite the source caption calling the model serial. End diagram.

Nodes

  1. Node 1: world w sub one¬p\mFalse{p}sourcep\mSat{{}}{\Box p}sourcew1w_1source
  2. Node 2: world w sub twop\mTrue{p}sourcep\mSat{{}}{\Box p}sourcew2w_2source
  3. Node 3: world w sub three¬p\mFalse{p}sourcep\mSat{{}}{\Box p}sourcew3w_3source
  4. Node 4: world w sub fourp\mTrue{p}sourcep\mSat/{{}}{\Box p}sourcew4w_4source
  5. Node 5: world w sub five¬p\mFalse{p}sourcep\mSat/{{}}{\Box p}sourcew5w_5source

Edges

  1. Edge 1: w1 to w2; accessibility relation R.
  2. Edge 2: w3 to w4; accessibility relation R.
  3. Edge 3: w4 to w4; accessibility relation R; loop.
  4. Edge 4: w4 to w5; accessibility relation R.
  5. Edge 5: w5 to w4; accessibility relation R.
  6. Edge 6: w5 to w5; accessibility relation R; loop.
source 27

captionA serial and euclidean model.

Figure of the filtered euclidean example

The outer figure contains the four-class filtration graph. The inner diagram records the merged class of w one and w three, every printed valuation and modal status, all arrows, and both loops.

Source transcription

[ht] centering

Four-class filtration source graph

Four-node filtration graph with the first and third source worlds merged. Nodes in source order. Source world node merged class w sub one and w sub three. Valuation label p is false. Printed node annotation the equivalence class of world w sub one equals the equivalence class of world w sub three. Modal-status label necessarily p is true at this displayed world. Printed node label the equivalence class of world w sub one. Source world node class w sub two. Valuation label p is true. Modal-status label necessarily p is true at this displayed world. Printed node label the equivalence class of world w sub two. Source world node class w sub four. Valuation label p is true. Modal-status label necessarily p is false at this displayed world. Printed node label the equivalence class of world w sub four. Source world node class w sub five. Valuation label p is false. Modal-status label necessarily p is false at this displayed world. Printed node label the equivalence class of world w sub five. Directed edges in source order. Directed accessibility edge from merged class w sub one and w sub three to class w sub two. Directed accessibility edge from merged class w sub one and w sub three to class w sub four. Directed accessibility loop at class w sub four. Directed accessibility edge from class w sub four to class w sub five. Directed accessibility edge from class w sub five to class w sub four. Directed accessibility loop at class w sub five. Omissions and continuation marks. The edge list is exactly the printed filtration. No unprinted double arrows or other euclidean repair edges from the surrounding prose are added. End diagram.

Nodes

  1. Node 1: the equivalence class of world w sub one¬p\mFalse{p}source[w1]=[w3][w_1]=[w_3]sourcep\mSat{{}}{\Box p}source[w1][w_1]source
  2. Node 2: the equivalence class of world w sub twop\mTrue{p}sourcep\mSat{{}}{\Box p}source[w2][w_2]source
  3. Node 3: the equivalence class of world w sub fourp\mTrue{p}sourcep\mSat/{{}}{\Box p}source[w4][w_4]source
  4. Node 4: the equivalence class of world w sub five¬p\mFalse{p}sourcep\mSat/{{}}{\Box p}source[w5][w_5]source

Edges

  1. Edge 1: w1 to w2; accessibility relation R star.
  2. Edge 2: w1 to w4; accessibility relation R star.
  3. Edge 3: w4 to w4; accessibility relation R star; loop.
  4. Edge 4: w4 to w5; accessibility relation R star.
  5. Edge 5: w5 to w4; accessibility relation R star.
  6. Edge 6: w5 to w5; accessibility relation R star; loop.
source 60

captionThe filtration of the model in the displayed source model described as serial and euclidean.

In particular, to obtain a euclidean filtration it is not enough to consider filtrations through arbitrary Γ\Gammasource's closed under sub-formulas. Instead we need to consider sets Γ\Gammasource that are modally closed (see the definition of a modally closed set of formulas). Such sets of sentences are infinite, and therefore do not immediately yield a finite model property or the decidability of the corresponding system.

Coarsest filtrations through modally closed sets

Let Γ\Gammasource be modally closed, M=W,R,V\mModel{M}=\tuple{W,R,V}source, and M*=W*,R*,V*\mModel{M^*} = \tuple{W^*,R^*,V^*}source be a coarsest filtration of M\mModel{M}source.

  1. If M\mModel{M}source is symmetric, so is M*\mModel{M^*}source.

  2. If M\mModel{M}source is transitive, so is M*\mModel{M^*}source.

  3. If M\mModel{M}source is euclidean, so is M*\mModel{M^*}source.

Proof

  1. If M*\mModel{M^*}source is a coarsest filtration, then by definition R*[u][v]R^*[u][v]source holds if and only if C1(u,v)C_1(u,v)source. For transitivity, suppose C1(u,v)C_1(u,v)source and C1(v,w)C_1(v,w)source; we have to show C1(u,w)C_1(u,w)source. Suppose MA[u]\mSat{M}{\Box !A}[u]source; then MA[u]\mSat{M}{\Box\Box!A}[u]source since Ax4 is valid in all transitive models; since AΓ\Box\Box!A \in \Gammasource by closure, also by C1(u,v)C_1(u,v)source, MA[v]\mSat{M}{\Box!A}[v]source and by C1(v,w)C_1(v,w)source, also MA[w]\mSat{M}{!A}[w]source. Suppose MA[w]\mSat{M}{!A}[w]source; then MA[v]\mSat{M}{\Diamond !A}[v]source by C1(v,w)C_1(v,w)source, since AΓ\Diamond !A \in \Gammasource by modal closure. By C1(u,v)C_1(u,v)source, we get MA[u]\mSat{M}{\Diamond\Diamond !A}[u]source since AΓ\Diamond\Diamond!A \in \Gammasource by modal closure. Since 4\Ax{4}_\Diamondsource is valid in all transitive models, MA[u]\mSat{M}{\Diamond!A}[u]source.

  2. Exercise. Use the fact that both Ax5 and 5\Ax{5_\Diamond}source are valid in all euclidean models.

  3. Exercise. Use the fact that AxB and B\Ax{B_\Diamond}source are valid in all symmetric models.

Exercise completing modally closed filtration cases

Complete the proof of the theorem on coarsest filtrations through modally closed sets.

Source disclosures