Contents 1. Introduction 2. (Un)decidability on modal MTL logics - - PowerPoint PPT Presentation

contents
SMART_READER_LITE
LIVE PREVIEW

Contents 1. Introduction 2. (Un)decidability on modal MTL logics - - PowerPoint PPT Presentation

Introduction (Un)decidability on modal MTL logics Undecidability of some modal MTL logics (formerly product logics) Amanda Vidal Institute of Computer Science, Czech Academy of Sciences September 6, 2016 1 / 19 Introduction (Un)decidability


slide-1
SLIDE 1

Introduction (Un)decidability on modal MTL logics

Undecidability of some modal MTL logics (formerly

product logics)

Amanda Vidal

Institute of Computer Science, Czech Academy of Sciences

September 6, 2016

1 / 19

slide-2
SLIDE 2

Introduction (Un)decidability on modal MTL logics

Contents

  • 1. Introduction
  • 2. (Un)decidability on modal MTL logics

Reducing to PCP

The Global modal logic case The Local modal logic case

2 / 19

slide-3
SLIDE 3

Introduction (Un)decidability on modal MTL logics

Contents

  • 1. Introduction
  • 2. (Un)decidability on modal MTL logics

Reducing to PCP

The Global modal logic case The Local modal logic case

3 / 19

slide-4
SLIDE 4

Introduction (Un)decidability on modal MTL logics

Introduction

◮ Many normal (classical) modal logics: finite model property +

finite axiomatizability ⇒ decidability

4 / 19

slide-5
SLIDE 5

Introduction (Un)decidability on modal MTL logics

Introduction

◮ Many normal (classical) modal logics: finite model property +

finite axiomatizability ⇒ decidability

◮ many-valued cases: ? no (usual) FMP or (known) R.E

axiomatization...for instance

4 / 19

slide-6
SLIDE 6

Introduction (Un)decidability on modal MTL logics

Introduction

◮ Many normal (classical) modal logics: finite model property +

finite axiomatizability ⇒ decidability

◮ many-valued cases: ? no (usual) FMP or (known) R.E

axiomatization...for instance

◮ Validity in the expansion of Gödel logic with modal operators

does not enjoy FMP with the usual Kripke semantics, but it is decidable [Metcalfe et.al.]

4 / 19

slide-7
SLIDE 7

Introduction (Un)decidability on modal MTL logics

Introduction

◮ Many normal (classical) modal logics: finite model property +

finite axiomatizability ⇒ decidability

◮ many-valued cases: ? no (usual) FMP or (known) R.E

axiomatization...for instance

◮ Validity in the expansion of Gödel logic with modal operators

does not enjoy FMP with the usual Kripke semantics, but it is decidable [Metcalfe et.al.]

◮ Similar concerning validity and >0-sat in FDL (multi-modal

variation) over Product logic [Cerami et. al]

4 / 19

slide-8
SLIDE 8

Introduction (Un)decidability on modal MTL logics

Introduction

◮ Many normal (classical) modal logics: finite model property +

finite axiomatizability ⇒ decidability

◮ many-valued cases: ? no (usual) FMP or (known) R.E

axiomatization...for instance

◮ Validity in the expansion of Gödel logic with modal operators

does not enjoy FMP with the usual Kripke semantics, but it is decidable [Metcalfe et.al.]

◮ Similar concerning validity and >0-sat in FDL (multi-modal

variation) over Product logic [Cerami et. al]

◮ The previous case with involutive negation or allowing GCI

(some globally valid formulas) is undecidable [Baader et.al]

4 / 19

slide-9
SLIDE 9

Introduction (Un)decidability on modal MTL logics

Introduction

◮ Many normal (classical) modal logics: finite model property +

finite axiomatizability ⇒ decidability

◮ many-valued cases: ? no (usual) FMP or (known) R.E

axiomatization...for instance

◮ Validity in the expansion of Gödel logic with modal operators

does not enjoy FMP with the usual Kripke semantics, but it is decidable [Metcalfe et.al.]

◮ Similar concerning validity and >0-sat in FDL (multi-modal

variation) over Product logic [Cerami et. al]

◮ The previous case with involutive negation or allowing GCI

(some globally valid formulas) is undecidable [Baader et.al]

◮ ... 4 / 19

slide-10
SLIDE 10

Introduction (Un)decidability on modal MTL logics

MTL Kripke-models

A = A, ⊙, ⇒, min, 1, 0, a complete MTL algebra (conm. integral bounded prelinear residuated lattices = algebras in the variety generated by all left-continuous t-noms). Language: &, ∧, →, 0 plus two unary (modal) symbols (✷, ✸)

5 / 19

slide-11
SLIDE 11

Introduction (Un)decidability on modal MTL logics

MTL Kripke-models

A = A, ⊙, ⇒, min, 1, 0, a complete MTL algebra (conm. integral bounded prelinear residuated lattices = algebras in the variety generated by all left-continuous t-noms). Language: &, ∧, →, 0 plus two unary (modal) symbols (✷, ✸)

Definition

A (crisp) A Kripke model M is a tripla W , R, e where:

◮ R ⊆ W × W (Rus stands for u, s ∈ R) ◮ e : W × Var → A uniquelly extended by:

◮ e(u, ϕ&ψ) = e(u, ϕ) ⊙ e(u, ψ);

e(u, ϕ → ψ) = e(u, ϕ) ⇒ e(u, ψ); e(u, ϕ ∧ ψ) = min{e(u, ϕ), e(u, ψ)}; e(e, 0) = 0

◮ e(u, ✷ϕ) = inf {e(s, ϕ) : Rus} ◮ e(u, ✸ϕ) = sup{e(s, ϕ) : Rus} 5 / 19

slide-12
SLIDE 12

Introduction (Un)decidability on modal MTL logics

Modal MTL logics

C a class of complete MTL-algebras.

◮ (Global deduction): Γ C ϕ iff

[∀u ∈ W e(u, [Γ]) ⊆ {1}] implies [∀u ∈ W e(u, ϕ) = 1] for all A Kripke models M with A ∈ C. Γ f

C ϕ for denoting the same relation over finite (i.e., finite

W) Kripke models.

6 / 19

slide-13
SLIDE 13

Introduction (Un)decidability on modal MTL logics

Modal MTL logics

C a class of complete MTL-algebras.

◮ (Global deduction): Γ C ϕ iff

[∀u ∈ W e(u, [Γ]) ⊆ {1}] implies [∀u ∈ W e(u, ϕ) = 1] for all A Kripke models M with A ∈ C. Γ f

C ϕ for denoting the same relation over finite (i.e., finite

W) Kripke models.

◮ (Local deduction): Γ ⊢4C ϕ iff

∀u ∈ W [e(u, [Γ]) ⊆ {1} implies e(u, ϕ) = 1] for all transitive A Kripke models M with A ∈ C. Γ ⊢f

4C ϕ for denoting the same relation over finite transitive

Kripke models

6 / 19

slide-14
SLIDE 14

Introduction (Un)decidability on modal MTL logics

Contents

  • 1. Introduction
  • 2. (Un)decidability on modal MTL logics

Reducing to PCP

The Global modal logic case The Local modal logic case

7 / 19

slide-15
SLIDE 15

Introduction (Un)decidability on modal MTL logics

Undecidability results

For n < ω, a MTL-algebra is n-contractive iff it validates the equation xn → xn+1 = 1 A class of MTL-algebras is non contractive iff, for all n, it contains some non n-contractive algebra.

Theorem

Let C be a non contractive class of complete MTL-algebras. For arbitrary Γ ∪ {ϕ} the following are undecidable:

8 / 19

slide-16
SLIDE 16

Introduction (Un)decidability on modal MTL logics

Undecidability results

For n < ω, a MTL-algebra is n-contractive iff it validates the equation xn → xn+1 = 1 A class of MTL-algebras is non contractive iff, for all n, it contains some non n-contractive algebra.

Theorem

Let C be a non contractive class of complete MTL-algebras. For arbitrary Γ ∪ {ϕ} the following are undecidable:

  • 1. Γ C ϕ
  • 2. Γ f

C ϕ (global deduction)

8 / 19

slide-17
SLIDE 17

Introduction (Un)decidability on modal MTL logics

Undecidability results

For n < ω, a MTL-algebra is n-contractive iff it validates the equation xn → xn+1 = 1 A class of MTL-algebras is non contractive iff, for all n, it contains some non n-contractive algebra.

Theorem

Let C be a non contractive class of complete MTL-algebras. For arbitrary Γ ∪ {ϕ} the following are undecidable:

  • 1. Γ C ϕ
  • 2. Γ f

C ϕ (global deduction)

  • 3. Γ ⊢4C ϕ
  • 4. Γ ⊢f

4C ϕ (local deduction in transitive frames)

8 / 19

slide-18
SLIDE 18

Introduction (Un)decidability on modal MTL logics

Post Correspondence Problem

An instance of the PCP is a list of pairs v1, w1 . . . vn, wn where vi, wi are numbers in base s ≥ 2.

9 / 19

slide-19
SLIDE 19

Introduction (Un)decidability on modal MTL logics

Post Correspondence Problem

An instance of the PCP is a list of pairs v1, w1 . . . vn, wn where vi, wi are numbers in base s ≥ 2. It is undecidable whether there exist i1, . . . , ik such that vi1 · · · vik = wi1 · · · wik

9 / 19

slide-20
SLIDE 20

Introduction (Un)decidability on modal MTL logics

Post Correspondence Problem

An instance of the PCP is a list of pairs v1, w1 . . . vn, wn where vi, wi are numbers in base s ≥ 2. It is undecidable whether there exist i1, . . . , ik such that vi1 · · · vik = wi1 · · · wik

◮ a, b numbers in base s =

⇒ ab = a · sb + b, where b is the length of b (in base s).

9 / 19

slide-21
SLIDE 21

Introduction (Un)decidability on modal MTL logics

Post Correspondence Problem

An instance of the PCP is a list of pairs v1, w1 . . . vn, wn where vi, wi are numbers in base s ≥ 2. It is undecidable whether there exist i1, . . . , ik such that vi1 · · · vik = wi1 · · · wik

◮ a, b numbers in base s =

⇒ ab = a · sb + b, where b is the length of b (in base s).

◮ we can exploit the conjunction operation to express

concatenation (using powers over some y ”non-contractive”)

9 / 19

slide-22
SLIDE 22

Introduction (Un)decidability on modal MTL logics

The global modal logic case

Given a PCP instance P there is a finite set of formulas Γg(P) ∪ {ϕg} such that P is SAT ⇐ ⇒ Γg(P) C ϕg Moreover Γg(P) C ϕg ⇐ ⇒ Γg(P) f

C ϕg.

10 / 19

slide-23
SLIDE 23

Introduction (Un)decidability on modal MTL logics

The global modal logic case

Given a PCP instance P there is a finite set of formulas Γg(P) ∪ {ϕg} such that P is SAT ⇐ ⇒ Γg(P) C ϕg Moreover Γg(P) C ϕg ⇐ ⇒ Γg(P) f

C ϕg. ◮ Proving =

⇒ will not be hard (constructing a model using the solution of P).

10 / 19

slide-24
SLIDE 24

Introduction (Un)decidability on modal MTL logics

The global modal logic case

Given a PCP instance P there is a finite set of formulas Γg(P) ∪ {ϕg} such that P is SAT ⇐ ⇒ Γg(P) C ϕg Moreover Γg(P) C ϕg ⇐ ⇒ Γg(P) f

C ϕg. ◮ Proving =

⇒ will not be hard (constructing a model using the solution of P).

◮ Idea for ⇐

=: if Γg(P) ϕg then it happens in uk of a particular structure shaped like

  • uk
  • uk−1
  • u2

u1

10 / 19

slide-25
SLIDE 25

Introduction (Un)decidability on modal MTL logics

The global case: formulas

Variables used: V = {x, y, z, v, w}. y, z, are control variables; x stores information on the index of the added word; v, w store information on the concatenation.

11 / 19

slide-26
SLIDE 26

Introduction (Un)decidability on modal MTL logics

The global case: formulas

Variables used: V = {x, y, z, v, w}. y, z, are control variables; x stores information on the index of the added word; v, w store information on the concatenation. Formulas of Γg(P):

◮ (¬✷0) → (✷p ↔ ✸p) for each p ∈ V:

11 / 19

slide-27
SLIDE 27

Introduction (Un)decidability on modal MTL logics

The global case: formulas

Variables used: V = {x, y, z, v, w}. y, z, are control variables; x stores information on the index of the added word; v, w store information on the concatenation. Formulas of Γg(P):

◮ (¬✷0) → (✷p ↔ ✸p) for each p ∈ V:

Lemma

If Γg(P) C ψ (for arbitrary ψ in V) then there is a C Kripke model M with W = {ui : i ∈ ω} or W = {ui : i ≤ k} and R = {ui, ui+1} such that

◮ M is a model for Γg(P) and ◮ e(u1, ψ) < 1 11 / 19

slide-28
SLIDE 28

Introduction (Un)decidability on modal MTL logics

The global case: formulas

◮ 1≤i≤n(x ↔ zi):

12 / 19

slide-29
SLIDE 29

Introduction (Un)decidability on modal MTL logics

The global case: formulas

◮ 1≤i≤n(x ↔ zi): at each world u, x = αi z for some 1 ≤ i ≤ n.

12 / 19

slide-30
SLIDE 30

Introduction (Un)decidability on modal MTL logics

The global case: formulas

◮ 1≤i≤n(x ↔ zi): at each world u, x = αi z for some 1 ≤ i ≤ n.

idea: if e(u, x) = αi

z, the number added in the concatenation

(to v and w) is the one indexed by i.

12 / 19

slide-31
SLIDE 31

Introduction (Un)decidability on modal MTL logics

The global case: formulas

◮ 1≤i≤n(x ↔ zi): at each world u, x = αi z for some 1 ≤ i ≤ n.

idea: if e(u, x) = αi

z, the number added in the concatenation

(to v and w) is the one indexed by i.

◮ (x ↔ zi) → (v ↔ (✷v)svi &yvi) for each 1 ≤ i ≤ n:

12 / 19

slide-32
SLIDE 32

Introduction (Un)decidability on modal MTL logics

The global case: formulas

◮ 1≤i≤n(x ↔ zi): at each world u, x = αi z for some 1 ≤ i ≤ n.

idea: if e(u, x) = αi

z, the number added in the concatenation

(to v and w) is the one indexed by i.

◮ (x ↔ zi) → (v ↔ (✷v)svi &yvi) for each 1 ≤ i ≤ n:

(information on the concatenation of vs)

12 / 19

slide-33
SLIDE 33

Introduction (Un)decidability on modal MTL logics

The global case: formulas

◮ 1≤i≤n(x ↔ zi): at each world u, x = αi z for some 1 ≤ i ≤ n.

idea: if e(u, x) = αi

z, the number added in the concatenation

(to v and w) is the one indexed by i.

◮ (x ↔ zi) → (v ↔ (✷v)svi &yvi) for each 1 ≤ i ≤ n:

(information on the concatenation of vs)

◮ (x ↔ zi) → (w ↔ (✷w)swi &ywi) for each 1 ≤ i ≤ n: (as

above for ws)

12 / 19

slide-34
SLIDE 34

Introduction (Un)decidability on modal MTL logics

The global case: formulas

◮ 1≤i≤n(x ↔ zi): at each world u, x = αi z for some 1 ≤ i ≤ n.

idea: if e(u, x) = αi

z, the number added in the concatenation

(to v and w) is the one indexed by i.

◮ (x ↔ zi) → (v ↔ (✷v)svi &yvi) for each 1 ≤ i ≤ n:

(information on the concatenation of vs)

◮ (x ↔ zi) → (w ↔ (✷w)swi &ywi) for each 1 ≤ i ≤ n: (as

above for ws) Let ϕg = (v ↔ w) → ((v → v&y) ∨ (w → w&y) ∨ (zn−1 → zn)).

12 / 19

slide-35
SLIDE 35

Introduction (Un)decidability on modal MTL logics

The global case: main result

Lemma

Let M with W = {ui : 1 ≤ i ≤ κ} and R = {ui+1, ui : 1 ≤ i < κ} be a model of Γg(P) such that e(uκ, ϕg) < 1. Then

  • 1. κ < ω (i.e, the model is finite)

13 / 19

slide-36
SLIDE 36

Introduction (Un)decidability on modal MTL logics

The global case: main result

Lemma

Let M with W = {ui : 1 ≤ i ≤ κ} and R = {ui+1, ui : 1 ≤ i < κ} be a model of Γg(P) such that e(uκ, ϕg) < 1. Then

  • 1. κ < ω (i.e, the model is finite)

e(uκ, v) = infi≤ν αi

y for some ν ≤ ω (same for w and

some λ).

13 / 19

slide-37
SLIDE 37

Introduction (Un)decidability on modal MTL logics

The global case: main result

Lemma

Let M with W = {ui : 1 ≤ i ≤ κ} and R = {ui+1, ui : 1 ≤ i < κ} be a model of Γg(P) such that e(uκ, ϕg) < 1. Then

  • 1. κ < ω (i.e, the model is finite)

e(uκ, v) = infi≤ν αi

y for some ν ≤ ω (same for w and

some λ). since e(uk, v → v&y) < 1 (and sim. for w) then ν, λ < ω and the model is of finite depth.

13 / 19

slide-38
SLIDE 38

Introduction (Un)decidability on modal MTL logics

The global case: main result

Lemma

Let M with W = {ui : 1 ≤ i ≤ κ} and R = {ui+1, ui : 1 ≤ i < κ} be a model of Γg(P) such that e(uκ, ϕg) < 1. Then

  • 1. κ < ω (i.e, the model is finite)

e(uκ, v) = infi≤ν αi

y for some ν ≤ ω (same for w and

some λ). since e(uk, v → v&y) < 1 (and sim. for w) then ν, λ < ω and the model is of finite depth.

  • 2. αn

z < ... < αz (determining indexes from 1 to n)

13 / 19

slide-39
SLIDE 39

Introduction (Un)decidability on modal MTL logics

The global case: main result

Lemma

Let M with W = {ui : 1 ≤ i ≤ κ} and R = {ui+1, ui : 1 ≤ i < κ} be a model of Γg(P) such that e(uκ, ϕg) < 1. Then

  • 1. κ < ω (i.e, the model is finite)

e(uκ, v) = infi≤ν αi

y for some ν ≤ ω (same for w and

some λ). since e(uk, v → v&y) < 1 (and sim. for w) then ν, λ < ω and the model is of finite depth.

  • 2. αn

z < ... < αz (determining indexes from 1 to n)

follows from e(uκ, zn) < e(uκ, zn−1)

13 / 19

slide-40
SLIDE 40

Introduction (Un)decidability on modal MTL logics

The global case: main result

  • 3. for all 1 ≤ j ≤ κ, e(uj, v) = α

vi1···vij y

and e(uj, w) = α

wi1···wij y

for e(uj, x) = αij

z for 1 ≤ j ≤ k.

14 / 19

slide-41
SLIDE 41

Introduction (Un)decidability on modal MTL logics

The global case: main result

  • 3. for all 1 ≤ j ≤ κ, e(uj, v) = α

vi1···vij y

and e(uj, w) = α

wi1···wij y

for e(uj, x) = αij

z for 1 ≤ j ≤ k.

provable by induction in j.

14 / 19

slide-42
SLIDE 42

Introduction (Un)decidability on modal MTL logics

The global case: main result

  • 3. for all 1 ≤ j ≤ κ, e(uj, v) = α

vi1···vij y

and e(uj, w) = α

wi1···wij y

for e(uj, x) = αij

z for 1 ≤ j ≤ k.

provable by induction in j.

  • 4. let a = max{vi1 · · · viκ, wi1 · · · wiκ}. For any 1 ≤ b < c ≤ a it

holds αc

y < αb y.

14 / 19

slide-43
SLIDE 43

Introduction (Un)decidability on modal MTL logics

The global case: main result

  • 3. for all 1 ≤ j ≤ κ, e(uj, v) = α

vi1···vij y

and e(uj, w) = α

wi1···wij y

for e(uj, x) = αij

z for 1 ≤ j ≤ k.

provable by induction in j.

  • 4. let a = max{vi1 · · · viκ, wi1 · · · wiκ}. For any 1 ≤ b < c ≤ a it

holds αc

y < αb y.

it follows from α

vi1···viκ+1 y

< α

vi1···viκ y

, which holds from previous point and e(uκ, v&y) < e(uκ, v) (same for w).

14 / 19

slide-44
SLIDE 44

Introduction (Un)decidability on modal MTL logics

The global case: main result

  • 3. for all 1 ≤ j ≤ κ, e(uj, v) = α

vi1···vij y

and e(uj, w) = α

wi1···wij y

for e(uj, x) = αij

z for 1 ≤ j ≤ k.

provable by induction in j.

  • 4. let a = max{vi1 · · · viκ, wi1 · · · wiκ}. For any 1 ≤ b < c ≤ a it

holds αc

y < αb y.

it follows from α

vi1···viκ+1 y

< α

vi1···viκ y

, which holds from previous point and e(uκ, v&y) < e(uκ, v) (same for w).

  • 5. e(uκ, v) = e(uκ, w) (so vi1 · · · viκ = wi1 · · · wiκ)

14 / 19

slide-45
SLIDE 45

Introduction (Un)decidability on modal MTL logics

The global case: main result

  • 3. for all 1 ≤ j ≤ κ, e(uj, v) = α

vi1···vij y

and e(uj, w) = α

wi1···wij y

for e(uj, x) = αij

z for 1 ≤ j ≤ k.

provable by induction in j.

  • 4. let a = max{vi1 · · · viκ, wi1 · · · wiκ}. For any 1 ≤ b < c ≤ a it

holds αc

y < αb y.

it follows from α

vi1···viκ+1 y

< α

vi1···viκ y

, which holds from previous point and e(uκ, v&y) < e(uκ, v) (same for w).

  • 5. e(uκ, v) = e(uκ, w) (so vi1 · · · viκ = wi1 · · · wiκ)
  • therwise, e(uκ, v ↔ w) ≤ αy and we know e(uκ, v →

v&y) ≥ αy (contradicting e(uκ, ϕg) < 1).

14 / 19

slide-46
SLIDE 46

Introduction (Un)decidability on modal MTL logics

From P to a model and back

◮ If Γg(P) (f ) C

ϕg in uk of a model M as the one from before we can naturally get a solution for P.

15 / 19

slide-47
SLIDE 47

Introduction (Un)decidability on modal MTL logics

From P to a model and back

◮ If Γg(P) (f ) C

ϕg in uk of a model M as the one from before we can naturally get a solution for P.

◮ If i1, . . . , ik is a solution for P, then Γg(P) (f ) C

ϕg in uk of the model M = {u1, . . . , uk}, {uk, uk−1, ..., u2, u1}, e with

15 / 19

slide-48
SLIDE 48

Introduction (Un)decidability on modal MTL logics

From P to a model and back

◮ If Γg(P) (f ) C

ϕg in uk of a model M as the one from before we can naturally get a solution for P.

◮ If i1, . . . , ik is a solution for P, then Γg(P) (f ) C

ϕg in uk of the model M = {u1, . . . , uk}, {uk, uk−1, ..., u2, u1}, e with

◮ e(u, y) = αy ∈ A(∈ C) such that αy (and so, A) is non

r-contractive for r depending on n and vi1 · · · vik,

15 / 19

slide-49
SLIDE 49

Introduction (Un)decidability on modal MTL logics

From P to a model and back

◮ If Γg(P) (f ) C

ϕg in uk of a model M as the one from before we can naturally get a solution for P.

◮ If i1, . . . , ik is a solution for P, then Γg(P) (f ) C

ϕg in uk of the model M = {u1, . . . , uk}, {uk, uk−1, ..., u2, u1}, e with

◮ e(u, y) = αy ∈ A(∈ C) such that αy (and so, A) is non

r-contractive for r depending on n and vi1 · · · vik,

◮ e(uj, v) = α

vi1···vij y

(analogously for w),

15 / 19

slide-50
SLIDE 50

Introduction (Un)decidability on modal MTL logics

From P to a model and back

◮ If Γg(P) (f ) C

ϕg in uk of a model M as the one from before we can naturally get a solution for P.

◮ If i1, . . . , ik is a solution for P, then Γg(P) (f ) C

ϕg in uk of the model M = {u1, . . . , uk}, {uk, uk−1, ..., u2, u1}, e with

◮ e(u, y) = αy ∈ A(∈ C) such that αy (and so, A) is non

r-contractive for r depending on n and vi1 · · · vik,

◮ e(uj, v) = α

vi1···vij y

(analogously for w),

◮ e(ui, z) = αm

y with m depending on vi1 · · · vik and wi1 · · · wik,

(αm

y = min1≤j≤k α vi1···vik y

↔ α

vi1···vik−1·vj y

)

15 / 19

slide-51
SLIDE 51

Introduction (Un)decidability on modal MTL logics

From P to a model and back

◮ If Γg(P) (f ) C

ϕg in uk of a model M as the one from before we can naturally get a solution for P.

◮ If i1, . . . , ik is a solution for P, then Γg(P) (f ) C

ϕg in uk of the model M = {u1, . . . , uk}, {uk, uk−1, ..., u2, u1}, e with

◮ e(u, y) = αy ∈ A(∈ C) such that αy (and so, A) is non

r-contractive for r depending on n and vi1 · · · vik,

◮ e(uj, v) = α

vi1···vij y

(analogously for w),

◮ e(ui, z) = αm

y with m depending on vi1 · · · vik and wi1 · · · wik,

(αm

y = min1≤j≤k α vi1···vik y

↔ α

vi1···vik−1·vj y

)

◮ e(uj, x) = αij

z ( observe e(uj, x ↔ zr) for 1 ≤ r ≤ n is either 1

(if r = ij) or is ≤ αz).

15 / 19

slide-52
SLIDE 52

Introduction (Un)decidability on modal MTL logics

The local modal logic case

In a similar fashion as before we can define a finite set ΓL(P) ∪ {ϕL} (in the same V) such that P is SAT ⇐ ⇒ ΓL(P) ⊢4C ϕ and that ΓL(P) ⊢4C ϕL ⇐ ⇒ ΓL(P) ⊢f

4C ϕL.

16 / 19

slide-53
SLIDE 53

Introduction (Un)decidability on modal MTL logics

The local modal logic case

In a similar fashion as before we can define a finite set ΓL(P) ∪ {ϕL} (in the same V) such that P is SAT ⇐ ⇒ ΓL(P) ⊢4C ϕ and that ΓL(P) ⊢4C ϕL ⇐ ⇒ ΓL(P) ⊢f

4C ϕL.

We now work towards structures with the form

  • uk
  • uk−1
  • u2

u1

  • u0
  • 16 / 19
slide-54
SLIDE 54

Introduction (Un)decidability on modal MTL logics

The local modal logic case: some differences

ΓL(P) set of formulas: very similar to Γg(P) but

17 / 19

slide-55
SLIDE 55

Introduction (Un)decidability on modal MTL logics

The local modal logic case: some differences

ΓL(P) set of formulas: very similar to Γg(P) but

◮ ✷(¬✷0 → (✷p ↔ ✸p)) is only added for y, z variables:

17 / 19

slide-56
SLIDE 56

Introduction (Un)decidability on modal MTL logics

The local modal logic case: some differences

ΓL(P) set of formulas: very similar to Γg(P) but

◮ ✷(¬✷0 → (✷p ↔ ✸p)) is only added for y, z variables: in all

the successors y and z are constant.

17 / 19

slide-57
SLIDE 57

Introduction (Un)decidability on modal MTL logics

The local modal logic case: some differences

ΓL(P) set of formulas: very similar to Γg(P) but

◮ ✷(¬✷0 → (✷p ↔ ✸p)) is only added for y, z variables: in all

the successors y and z are constant.

◮ ✸✷0, ✷(✷0&x) ↔ ✸(✷0&x) are added:

17 / 19

slide-58
SLIDE 58

Introduction (Un)decidability on modal MTL logics

The local modal logic case: some differences

ΓL(P) set of formulas: very similar to Γg(P) but

◮ ✷(¬✷0 → (✷p ↔ ✸p)) is only added for y, z variables: in all

the successors y and z are constant.

◮ ✸✷0, ✷(✷0&x) ↔ ✸(✷0&x) are added: there is some world

with no successors, and in all them x is constant (so it will be v, w)

17 / 19

slide-59
SLIDE 59

Introduction (Un)decidability on modal MTL logics

The local modal logic case: some differences

ΓL(P) set of formulas: very similar to Γg(P) but

◮ ✷(¬✷0 → (✷p ↔ ✸p)) is only added for y, z variables: in all

the successors y and z are constant.

◮ ✸✷0, ✷(✷0&x) ↔ ✸(✷0&x) are added: there is some world

with no successors, and in all them x is constant (so it will be v, w)

◮ Formulas determining values of x, v, w are the ones from

Γg(P) closed by a ✷.

17 / 19

slide-60
SLIDE 60

Introduction (Un)decidability on modal MTL logics

The local modal logic case: some differences

ΓL(P) set of formulas: very similar to Γg(P) but

◮ ✷(¬✷0 → (✷p ↔ ✸p)) is only added for y, z variables: in all

the successors y and z are constant.

◮ ✸✷0, ✷(✷0&x) ↔ ✸(✷0&x) are added: there is some world

with no successors, and in all them x is constant (so it will be v, w)

◮ Formulas determining values of x, v, w are the ones from

Γg(P) closed by a ✷.

◮ ✷(✷(v&w) → (✷v&✷w)):

17 / 19

slide-61
SLIDE 61

Introduction (Un)decidability on modal MTL logics

The local modal logic case: some differences

ΓL(P) set of formulas: very similar to Γg(P) but

◮ ✷(¬✷0 → (✷p ↔ ✸p)) is only added for y, z variables: in all

the successors y and z are constant.

◮ ✸✷0, ✷(✷0&x) ↔ ✸(✷0&x) are added: there is some world

with no successors, and in all them x is constant (so it will be v, w)

◮ Formulas determining values of x, v, w are the ones from

Γg(P) closed by a ✷.

◮ ✷(✷(v&w) → (✷v&✷w)): helps ensure the witness of ✷v

and ✷w coincides.

17 / 19

slide-62
SLIDE 62

Introduction (Un)decidability on modal MTL logics

The local modal logic case: some differences

ΓL(P) set of formulas: very similar to Γg(P) but

◮ ✷(¬✷0 → (✷p ↔ ✸p)) is only added for y, z variables: in all

the successors y and z are constant.

◮ ✸✷0, ✷(✷0&x) ↔ ✸(✷0&x) are added: there is some world

with no successors, and in all them x is constant (so it will be v, w)

◮ Formulas determining values of x, v, w are the ones from

Γg(P) closed by a ✷.

◮ ✷(✷(v&w) → (✷v&✷w)): helps ensure the witness of ✷v

and ✷w coincides. Let ϕL = ✷((v ↔ w) → ((v → v&y) ∨ (w → w&y) ∨ (v&w → v&w&y) ∨ (zn−1 → zn)))

17 / 19

slide-63
SLIDE 63

Introduction (Un)decidability on modal MTL logics

The local modal logic case: some differences

ΓL(P) set of formulas: very similar to Γg(P) but

◮ ✷(¬✷0 → (✷p ↔ ✸p)) is only added for y, z variables: in all

the successors y and z are constant.

◮ ✸✷0, ✷(✷0&x) ↔ ✸(✷0&x) are added: there is some world

with no successors, and in all them x is constant (so it will be v, w)

◮ Formulas determining values of x, v, w are the ones from

Γg(P) closed by a ✷.

◮ ✷(✷(v&w) → (✷v&✷w)): helps ensure the witness of ✷v

and ✷w coincides. Let ϕL = ✷((v ↔ w) → ((v → v&y) ∨ (w → w&y) ∨ (v&w → v&w&y) ∨ (zn−1 → zn))) the new part is linked to the uniqueness in the witness of ✷v, ✷w.

17 / 19

slide-64
SLIDE 64

Introduction (Un)decidability on modal MTL logics

The local modal logic case: procedure differences

◮ e(uj, y) = αy ∈ A and e(uj, z) = αz ∈ A for each 1 ≤ j ≤ k

are poved as before,

18 / 19

slide-65
SLIDE 65

Introduction (Un)decidability on modal MTL logics

The local modal logic case: procedure differences

◮ e(uj, y) = αy ∈ A and e(uj, z) = αz ∈ A for each 1 ≤ j ≤ k

are poved as before,

◮ If ΓL(P) ⊢4C ϕL, to check the desired completeness wrt the

depicted structures we show

  • 1. The model is finite: finite depth as before, finite width based
  • n the finite possible values for v and w,

18 / 19

slide-66
SLIDE 66

Introduction (Un)decidability on modal MTL logics

The local modal logic case: procedure differences

◮ e(uj, y) = αy ∈ A and e(uj, z) = αz ∈ A for each 1 ≤ j ≤ k

are poved as before,

◮ If ΓL(P) ⊢4C ϕL, to check the desired completeness wrt the

depicted structures we show

  • 1. The model is finite: finite depth as before, finite width based
  • n the finite possible values for v and w,
  • 2. The worlds witnessing ✷v and ✷w coincide (using the new

formula distributing ✷ over &)

18 / 19

slide-67
SLIDE 67

Introduction (Un)decidability on modal MTL logics

The local modal logic case: procedure differences

◮ e(uj, y) = αy ∈ A and e(uj, z) = αz ∈ A for each 1 ≤ j ≤ k

are poved as before,

◮ If ΓL(P) ⊢4C ϕL, to check the desired completeness wrt the

depicted structures we show

  • 1. The model is finite: finite depth as before, finite width based
  • n the finite possible values for v and w,
  • 2. The worlds witnessing ✷v and ✷w coincide (using the new

formula distributing ✷ over &)

The construction of a model M from a solution of P and viceversa are similar to the ones from the global case.

18 / 19

slide-68
SLIDE 68

Introduction (Un)decidability on modal MTL logics

Thank you!

19 / 19