Semantics: Residuated Frames Peter Jipsen, Chapman University, - - PowerPoint PPT Presentation

semantics residuated frames
SMART_READER_LITE
LIVE PREVIEW

Semantics: Residuated Frames Peter Jipsen, Chapman University, - - PowerPoint PPT Presentation

Semantics: Residuated Frames Peter Jipsen, Chapman University, Orange, California, USA joint work with Nikolaos Galatos, University of Denver, Colorado, USA ICLA, January, 2009 P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames


slide-1
SLIDE 1

Semantics: Residuated Frames

Peter Jipsen, Chapman University, Orange, California, USA joint work with Nikolaos Galatos, University of Denver, Colorado, USA ICLA, January, 2009

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 1 / 39

slide-2
SLIDE 2

Outline

Part I Universal Algebra Examples of residuated lattices Congruences and normal filters The lattice of subvarieties Varieties generated by positive universal classes Direct decompositions and poset products Part II Residuated Frames Decidability Poset products of residuated lattices

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 2 / 39

slide-3
SLIDE 3

Outline

Part I Universal Algebra Examples of residuated lattices Congruences and normal filters The lattice of subvarieties Varieties generated by positive universal classes Direct decompositions and poset products Part II Residuated Frames Decidability Poset products of residuated lattices

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 2 / 39

slide-4
SLIDE 4

Outline

Part I Universal Algebra Examples of residuated lattices Congruences and normal filters The lattice of subvarieties Varieties generated by positive universal classes Direct decompositions and poset products Part II Residuated Frames Decidability Poset products of residuated lattices

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 2 / 39

slide-5
SLIDE 5

Outline

Part I Universal Algebra Examples of residuated lattices Congruences and normal filters The lattice of subvarieties Varieties generated by positive universal classes Direct decompositions and poset products Part II Residuated Frames Decidability Poset products of residuated lattices

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 2 / 39

slide-6
SLIDE 6

Part II: Frames of residuated lattices

We now consider semantics for residuated lattices Kripke frames for modal logics are a useful tool for building counter models They also led to many interesting notions such as frame completeness, canonical frames, and correspondence results Aim: To present frames for arbitrary residuated lattices and connect them with the proof theory of substructural logic

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 3 / 39

slide-7
SLIDE 7

Part II: Frames of residuated lattices

We now consider semantics for residuated lattices Kripke frames for modal logics are a useful tool for building counter models They also led to many interesting notions such as frame completeness, canonical frames, and correspondence results Aim: To present frames for arbitrary residuated lattices and connect them with the proof theory of substructural logic

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 3 / 39

slide-8
SLIDE 8

Part II: Frames of residuated lattices

We now consider semantics for residuated lattices Kripke frames for modal logics are a useful tool for building counter models They also led to many interesting notions such as frame completeness, canonical frames, and correspondence results Aim: To present frames for arbitrary residuated lattices and connect them with the proof theory of substructural logic

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 3 / 39

slide-9
SLIDE 9

Part II: Frames of residuated lattices

We now consider semantics for residuated lattices Kripke frames for modal logics are a useful tool for building counter models They also led to many interesting notions such as frame completeness, canonical frames, and correspondence results Aim: To present frames for arbitrary residuated lattices and connect them with the proof theory of substructural logic

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 3 / 39

slide-10
SLIDE 10

Part II: Frames of residuated lattices

We now consider semantics for residuated lattices Kripke frames for modal logics are a useful tool for building counter models They also led to many interesting notions such as frame completeness, canonical frames, and correspondence results Aim: To present frames for arbitrary residuated lattices and connect them with the proof theory of substructural logic

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 3 / 39

slide-11
SLIDE 11

Galois connections and closure operators

For posets P, Q, maps f : P → Q, g : Q → P are a Galois connection if y ≤ f (x) ⇐ ⇒ x ≤ g(y), for all x ∈ P, y ∈ Q A map c : P → P is a closure operator if x ≤ y implies c(x) ≤ c(y), x ≤ c(x) and c(c(x)) = c(x) Exercise: If f , g are a Galois connection then c(x) = g(f (x)) is a closure

  • perator on P

A lattice frame is a structure W = (W , W ′, N) where W and W ′ are sets and N is a binary relation from W to W ′ E.g. If L is a lattice, WL = (L, L, ≤) is a lattice frame Let J(L) be the set completely join irreducibles and M(L) be the set completely meet irreducibles of L. Then L+ = (J(L), M(L), ≤) is a lattice frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 4 / 39

slide-12
SLIDE 12

Galois connections and closure operators

For posets P, Q, maps f : P → Q, g : Q → P are a Galois connection if y ≤ f (x) ⇐ ⇒ x ≤ g(y), for all x ∈ P, y ∈ Q A map c : P → P is a closure operator if x ≤ y implies c(x) ≤ c(y), x ≤ c(x) and c(c(x)) = c(x) Exercise: If f , g are a Galois connection then c(x) = g(f (x)) is a closure

  • perator on P

A lattice frame is a structure W = (W , W ′, N) where W and W ′ are sets and N is a binary relation from W to W ′ E.g. If L is a lattice, WL = (L, L, ≤) is a lattice frame Let J(L) be the set completely join irreducibles and M(L) be the set completely meet irreducibles of L. Then L+ = (J(L), M(L), ≤) is a lattice frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 4 / 39

slide-13
SLIDE 13

Galois connections and closure operators

For posets P, Q, maps f : P → Q, g : Q → P are a Galois connection if y ≤ f (x) ⇐ ⇒ x ≤ g(y), for all x ∈ P, y ∈ Q A map c : P → P is a closure operator if x ≤ y implies c(x) ≤ c(y), x ≤ c(x) and c(c(x)) = c(x) Exercise: If f , g are a Galois connection then c(x) = g(f (x)) is a closure

  • perator on P

A lattice frame is a structure W = (W , W ′, N) where W and W ′ are sets and N is a binary relation from W to W ′ E.g. If L is a lattice, WL = (L, L, ≤) is a lattice frame Let J(L) be the set completely join irreducibles and M(L) be the set completely meet irreducibles of L. Then L+ = (J(L), M(L), ≤) is a lattice frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 4 / 39

slide-14
SLIDE 14

Galois connections and closure operators

For posets P, Q, maps f : P → Q, g : Q → P are a Galois connection if y ≤ f (x) ⇐ ⇒ x ≤ g(y), for all x ∈ P, y ∈ Q A map c : P → P is a closure operator if x ≤ y implies c(x) ≤ c(y), x ≤ c(x) and c(c(x)) = c(x) Exercise: If f , g are a Galois connection then c(x) = g(f (x)) is a closure

  • perator on P

A lattice frame is a structure W = (W , W ′, N) where W and W ′ are sets and N is a binary relation from W to W ′ E.g. If L is a lattice, WL = (L, L, ≤) is a lattice frame Let J(L) be the set completely join irreducibles and M(L) be the set completely meet irreducibles of L. Then L+ = (J(L), M(L), ≤) is a lattice frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 4 / 39

slide-15
SLIDE 15

Galois connections and closure operators

For posets P, Q, maps f : P → Q, g : Q → P are a Galois connection if y ≤ f (x) ⇐ ⇒ x ≤ g(y), for all x ∈ P, y ∈ Q A map c : P → P is a closure operator if x ≤ y implies c(x) ≤ c(y), x ≤ c(x) and c(c(x)) = c(x) Exercise: If f , g are a Galois connection then c(x) = g(f (x)) is a closure

  • perator on P

A lattice frame is a structure W = (W , W ′, N) where W and W ′ are sets and N is a binary relation from W to W ′ E.g. If L is a lattice, WL = (L, L, ≤) is a lattice frame Let J(L) be the set completely join irreducibles and M(L) be the set completely meet irreducibles of L. Then L+ = (J(L), M(L), ≤) is a lattice frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 4 / 39

slide-16
SLIDE 16

Galois connections and closure operators

For posets P, Q, maps f : P → Q, g : Q → P are a Galois connection if y ≤ f (x) ⇐ ⇒ x ≤ g(y), for all x ∈ P, y ∈ Q A map c : P → P is a closure operator if x ≤ y implies c(x) ≤ c(y), x ≤ c(x) and c(c(x)) = c(x) Exercise: If f , g are a Galois connection then c(x) = g(f (x)) is a closure

  • perator on P

A lattice frame is a structure W = (W , W ′, N) where W and W ′ are sets and N is a binary relation from W to W ′ E.g. If L is a lattice, WL = (L, L, ≤) is a lattice frame Let J(L) be the set completely join irreducibles and M(L) be the set completely meet irreducibles of L. Then L+ = (J(L), M(L), ≤) is a lattice frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 4 / 39

slide-17
SLIDE 17

Lattice frames

For X ⊆ W and Y ⊆ W ′ we define the polarities X ⊲ = {b ∈ W ′ : x N b, for all x ∈ X} Y ⊳ = {a ∈ W : a N y, for all y ∈ Y } Exercise: The maps ⊲: P(W ) → P(W ′) and ⊳: P(W ′) → P(W ) form a Galois connection If γN(X) = X ⊲⊳ then γN : P(W ) → P(W ) is a closure operator

  • Lemma. If L = (L, ∧, ∨) is a lattice and γ is a closure operator on L, then

(γ[L], ∧, ∨γ) is a lattice where x ∨γ y = γ(x ∨ y)

  • Corollary. If W is a lattice frame then the Galois algebra

W+ = (γN[P(W )], ∩, ∪γN) is a complete lattice If L is a lattice, W+

L is the Dedekind-MacNeille completion of L and

x → {x}⊳ is an embedding

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 5 / 39

slide-18
SLIDE 18

Lattice frames

For X ⊆ W and Y ⊆ W ′ we define the polarities X ⊲ = {b ∈ W ′ : x N b, for all x ∈ X} Y ⊳ = {a ∈ W : a N y, for all y ∈ Y } Exercise: The maps ⊲: P(W ) → P(W ′) and ⊳: P(W ′) → P(W ) form a Galois connection If γN(X) = X ⊲⊳ then γN : P(W ) → P(W ) is a closure operator

  • Lemma. If L = (L, ∧, ∨) is a lattice and γ is a closure operator on L, then

(γ[L], ∧, ∨γ) is a lattice where x ∨γ y = γ(x ∨ y)

  • Corollary. If W is a lattice frame then the Galois algebra

W+ = (γN[P(W )], ∩, ∪γN) is a complete lattice If L is a lattice, W+

L is the Dedekind-MacNeille completion of L and

x → {x}⊳ is an embedding

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 5 / 39

slide-19
SLIDE 19

Lattice frames

For X ⊆ W and Y ⊆ W ′ we define the polarities X ⊲ = {b ∈ W ′ : x N b, for all x ∈ X} Y ⊳ = {a ∈ W : a N y, for all y ∈ Y } Exercise: The maps ⊲: P(W ) → P(W ′) and ⊳: P(W ′) → P(W ) form a Galois connection If γN(X) = X ⊲⊳ then γN : P(W ) → P(W ) is a closure operator

  • Lemma. If L = (L, ∧, ∨) is a lattice and γ is a closure operator on L, then

(γ[L], ∧, ∨γ) is a lattice where x ∨γ y = γ(x ∨ y)

  • Corollary. If W is a lattice frame then the Galois algebra

W+ = (γN[P(W )], ∩, ∪γN) is a complete lattice If L is a lattice, W+

L is the Dedekind-MacNeille completion of L and

x → {x}⊳ is an embedding

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 5 / 39

slide-20
SLIDE 20

Lattice frames

For X ⊆ W and Y ⊆ W ′ we define the polarities X ⊲ = {b ∈ W ′ : x N b, for all x ∈ X} Y ⊳ = {a ∈ W : a N y, for all y ∈ Y } Exercise: The maps ⊲: P(W ) → P(W ′) and ⊳: P(W ′) → P(W ) form a Galois connection If γN(X) = X ⊲⊳ then γN : P(W ) → P(W ) is a closure operator

  • Lemma. If L = (L, ∧, ∨) is a lattice and γ is a closure operator on L, then

(γ[L], ∧, ∨γ) is a lattice where x ∨γ y = γ(x ∨ y)

  • Corollary. If W is a lattice frame then the Galois algebra

W+ = (γN[P(W )], ∩, ∪γN) is a complete lattice If L is a lattice, W+

L is the Dedekind-MacNeille completion of L and

x → {x}⊳ is an embedding

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 5 / 39

slide-21
SLIDE 21

Lattice frames

For X ⊆ W and Y ⊆ W ′ we define the polarities X ⊲ = {b ∈ W ′ : x N b, for all x ∈ X} Y ⊳ = {a ∈ W : a N y, for all y ∈ Y } Exercise: The maps ⊲: P(W ) → P(W ′) and ⊳: P(W ′) → P(W ) form a Galois connection If γN(X) = X ⊲⊳ then γN : P(W ) → P(W ) is a closure operator

  • Lemma. If L = (L, ∧, ∨) is a lattice and γ is a closure operator on L, then

(γ[L], ∧, ∨γ) is a lattice where x ∨γ y = γ(x ∨ y)

  • Corollary. If W is a lattice frame then the Galois algebra

W+ = (γN[P(W )], ∩, ∪γN) is a complete lattice If L is a lattice, W+

L is the Dedekind-MacNeille completion of L and

x → {x}⊳ is an embedding

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 5 / 39

slide-22
SLIDE 22

Residuated frames

A residuated frame is a structure W = (W , W ′, N, ◦, , ) where W and W ′ are sets, N ⊆ W × W ′, ◦ ⊆ W 3, ⊆ W × W ′ × W and ⊆ W ′ × W × W such that for all x, y ∈ W , w ∈ W ′ (x ◦ y) N w ⇔ y N (x w) ⇔ x N (w y) Here x ◦ y = {z | (x, y, z) ∈ ◦} and similarly for , We also use X N y to abbreviate x N y for all x ∈ X and likewise for x N Y A ternary relation structure W = (W , ◦) is said to be associative if it satisfies (x ◦ y) ◦ z = x ◦ (y ◦ z), i.e., if it satisfies the following equivalence ∃u[(x, y, u) ∈ ◦ and (u, z, w) ∈ ◦] ⇐ ⇒ ∃v[(x, v, w) ∈ ◦ and (y, z, v) ∈ ◦] It is said to have a unit E ⊆ W if x ◦ E = {x} = E ◦ x, i.e., if ∃e ∈ E[(x, e, y) ∈ ◦] ⇐ ⇒ x = y ⇐ ⇒ ∃e ∈ E[(e, x, y) ∈ ◦]

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 6 / 39

slide-23
SLIDE 23

Residuated frames

A residuated frame is a structure W = (W , W ′, N, ◦, , ) where W and W ′ are sets, N ⊆ W × W ′, ◦ ⊆ W 3, ⊆ W × W ′ × W and ⊆ W ′ × W × W such that for all x, y ∈ W , w ∈ W ′ (x ◦ y) N w ⇔ y N (x w) ⇔ x N (w y) Here x ◦ y = {z | (x, y, z) ∈ ◦} and similarly for , We also use X N y to abbreviate x N y for all x ∈ X and likewise for x N Y A ternary relation structure W = (W , ◦) is said to be associative if it satisfies (x ◦ y) ◦ z = x ◦ (y ◦ z), i.e., if it satisfies the following equivalence ∃u[(x, y, u) ∈ ◦ and (u, z, w) ∈ ◦] ⇐ ⇒ ∃v[(x, v, w) ∈ ◦ and (y, z, v) ∈ ◦] It is said to have a unit E ⊆ W if x ◦ E = {x} = E ◦ x, i.e., if ∃e ∈ E[(x, e, y) ∈ ◦] ⇐ ⇒ x = y ⇐ ⇒ ∃e ∈ E[(e, x, y) ∈ ◦]

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 6 / 39

slide-24
SLIDE 24

Residuated frames

A residuated frame is a structure W = (W , W ′, N, ◦, , ) where W and W ′ are sets, N ⊆ W × W ′, ◦ ⊆ W 3, ⊆ W × W ′ × W and ⊆ W ′ × W × W such that for all x, y ∈ W , w ∈ W ′ (x ◦ y) N w ⇔ y N (x w) ⇔ x N (w y) Here x ◦ y = {z | (x, y, z) ∈ ◦} and similarly for , We also use X N y to abbreviate x N y for all x ∈ X and likewise for x N Y A ternary relation structure W = (W , ◦) is said to be associative if it satisfies (x ◦ y) ◦ z = x ◦ (y ◦ z), i.e., if it satisfies the following equivalence ∃u[(x, y, u) ∈ ◦ and (u, z, w) ∈ ◦] ⇐ ⇒ ∃v[(x, v, w) ∈ ◦ and (y, z, v) ∈ ◦] It is said to have a unit E ⊆ W if x ◦ E = {x} = E ◦ x, i.e., if ∃e ∈ E[(x, e, y) ∈ ◦] ⇐ ⇒ x = y ⇐ ⇒ ∃e ∈ E[(e, x, y) ∈ ◦]

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 6 / 39

slide-25
SLIDE 25

Residuated frames

A residuated frame is a structure W = (W , W ′, N, ◦, , ) where W and W ′ are sets, N ⊆ W × W ′, ◦ ⊆ W 3, ⊆ W × W ′ × W and ⊆ W ′ × W × W such that for all x, y ∈ W , w ∈ W ′ (x ◦ y) N w ⇔ y N (x w) ⇔ x N (w y) Here x ◦ y = {z | (x, y, z) ∈ ◦} and similarly for , We also use X N y to abbreviate x N y for all x ∈ X and likewise for x N Y A ternary relation structure W = (W , ◦) is said to be associative if it satisfies (x ◦ y) ◦ z = x ◦ (y ◦ z), i.e., if it satisfies the following equivalence ∃u[(x, y, u) ∈ ◦ and (u, z, w) ∈ ◦] ⇐ ⇒ ∃v[(x, v, w) ∈ ◦ and (y, z, v) ∈ ◦] It is said to have a unit E ⊆ W if x ◦ E = {x} = E ◦ x, i.e., if ∃e ∈ E[(x, e, y) ∈ ◦] ⇐ ⇒ x = y ⇐ ⇒ ∃e ∈ E[(e, x, y) ∈ ◦]

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 6 / 39

slide-26
SLIDE 26

Nuclei

A nucleus γ on a residuated lattice L is a closure operator on L such that γ(x)γ(y) ≤ γ(xy) (or γ(γ(x)γ(y)) = γ(xy)).

  • Theorem. Given a RL L = (L, ∧, ∨, ·, \, /, 1) and a nucleus on L, the

algebra Lγ = (Lγ, ∧, ∨γ, ·γ, \, /, γ(1)), is a residuated lattice, where x ·γ y = γ(x · y), x ∨γ y = γ(x ∨ y).

  • Theorem. For a frame W, γN is a nucleus on (P(W ), ∩, ∪, ◦, \, /, {1})
  • Corollary. If W is a residuated frame then the Galois algebra

W+ = (P(W ), ∩, ∪, ◦, \, /, 1)γN is a complete residuated lattice. Moreover, for WL, x → {x}⊳ is an embedding. If L is a RL, WL = (L, L, ≤, ·, \, /) is a residuated frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 7 / 39

slide-27
SLIDE 27

Nuclei

A nucleus γ on a residuated lattice L is a closure operator on L such that γ(x)γ(y) ≤ γ(xy) (or γ(γ(x)γ(y)) = γ(xy)).

  • Theorem. Given a RL L = (L, ∧, ∨, ·, \, /, 1) and a nucleus on L, the

algebra Lγ = (Lγ, ∧, ∨γ, ·γ, \, /, γ(1)), is a residuated lattice, where x ·γ y = γ(x · y), x ∨γ y = γ(x ∨ y).

  • Theorem. For a frame W, γN is a nucleus on (P(W ), ∩, ∪, ◦, \, /, {1})
  • Corollary. If W is a residuated frame then the Galois algebra

W+ = (P(W ), ∩, ∪, ◦, \, /, 1)γN is a complete residuated lattice. Moreover, for WL, x → {x}⊳ is an embedding. If L is a RL, WL = (L, L, ≤, ·, \, /) is a residuated frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 7 / 39

slide-28
SLIDE 28

Nuclei

A nucleus γ on a residuated lattice L is a closure operator on L such that γ(x)γ(y) ≤ γ(xy) (or γ(γ(x)γ(y)) = γ(xy)).

  • Theorem. Given a RL L = (L, ∧, ∨, ·, \, /, 1) and a nucleus on L, the

algebra Lγ = (Lγ, ∧, ∨γ, ·γ, \, /, γ(1)), is a residuated lattice, where x ·γ y = γ(x · y), x ∨γ y = γ(x ∨ y).

  • Theorem. For a frame W, γN is a nucleus on (P(W ), ∩, ∪, ◦, \, /, {1})
  • Corollary. If W is a residuated frame then the Galois algebra

W+ = (P(W ), ∩, ∪, ◦, \, /, 1)γN is a complete residuated lattice. Moreover, for WL, x → {x}⊳ is an embedding. If L is a RL, WL = (L, L, ≤, ·, \, /) is a residuated frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 7 / 39

slide-29
SLIDE 29

Nuclei

A nucleus γ on a residuated lattice L is a closure operator on L such that γ(x)γ(y) ≤ γ(xy) (or γ(γ(x)γ(y)) = γ(xy)).

  • Theorem. Given a RL L = (L, ∧, ∨, ·, \, /, 1) and a nucleus on L, the

algebra Lγ = (Lγ, ∧, ∨γ, ·γ, \, /, γ(1)), is a residuated lattice, where x ·γ y = γ(x · y), x ∨γ y = γ(x ∨ y).

  • Theorem. For a frame W, γN is a nucleus on (P(W ), ∩, ∪, ◦, \, /, {1})
  • Corollary. If W is a residuated frame then the Galois algebra

W+ = (P(W ), ∩, ∪, ◦, \, /, 1)γN is a complete residuated lattice. Moreover, for WL, x → {x}⊳ is an embedding. If L is a RL, WL = (L, L, ≤, ·, \, /) is a residuated frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 7 / 39

slide-30
SLIDE 30

Nuclei

A nucleus γ on a residuated lattice L is a closure operator on L such that γ(x)γ(y) ≤ γ(xy) (or γ(γ(x)γ(y)) = γ(xy)).

  • Theorem. Given a RL L = (L, ∧, ∨, ·, \, /, 1) and a nucleus on L, the

algebra Lγ = (Lγ, ∧, ∨γ, ·γ, \, /, γ(1)), is a residuated lattice, where x ·γ y = γ(x · y), x ∨γ y = γ(x ∨ y).

  • Theorem. For a frame W, γN is a nucleus on (P(W ), ∩, ∪, ◦, \, /, {1})
  • Corollary. If W is a residuated frame then the Galois algebra

W+ = (P(W ), ∩, ∪, ◦, \, /, 1)γN is a complete residuated lattice. Moreover, for WL, x → {x}⊳ is an embedding. If L is a RL, WL = (L, L, ≤, ·, \, /) is a residuated frame.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 7 / 39

slide-31
SLIDE 31

Frames of complete perfect lattices

A lattice L is perfect if every element is a join of elements of J(L) and a meet of elements of M(L) E.g. a Boolean algebra is perfect iff it is atomic For a perfect residuated lattice A, let A+ = (J(A), M(A), ≤, ◦, , , E) where x ◦ y = {z ∈ J(A) | z ≤ xy} and E = {z ∈ J(A) | z ≤ 1}

Theorem

A+ is a residuated frame and if A is complete then (A+)+ ∼ = A In particular, any finite lattice is complete and perfect So for finite residuated lattices, residuated frames give a compact representation analogous to atom structures for relation algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 8 / 39

slide-32
SLIDE 32

Frames of complete perfect lattices

A lattice L is perfect if every element is a join of elements of J(L) and a meet of elements of M(L) E.g. a Boolean algebra is perfect iff it is atomic For a perfect residuated lattice A, let A+ = (J(A), M(A), ≤, ◦, , , E) where x ◦ y = {z ∈ J(A) | z ≤ xy} and E = {z ∈ J(A) | z ≤ 1}

Theorem

A+ is a residuated frame and if A is complete then (A+)+ ∼ = A In particular, any finite lattice is complete and perfect So for finite residuated lattices, residuated frames give a compact representation analogous to atom structures for relation algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 8 / 39

slide-33
SLIDE 33

Frames of complete perfect lattices

A lattice L is perfect if every element is a join of elements of J(L) and a meet of elements of M(L) E.g. a Boolean algebra is perfect iff it is atomic For a perfect residuated lattice A, let A+ = (J(A), M(A), ≤, ◦, , , E) where x ◦ y = {z ∈ J(A) | z ≤ xy} and E = {z ∈ J(A) | z ≤ 1}

Theorem

A+ is a residuated frame and if A is complete then (A+)+ ∼ = A In particular, any finite lattice is complete and perfect So for finite residuated lattices, residuated frames give a compact representation analogous to atom structures for relation algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 8 / 39

slide-34
SLIDE 34

Frames of complete perfect lattices

A lattice L is perfect if every element is a join of elements of J(L) and a meet of elements of M(L) E.g. a Boolean algebra is perfect iff it is atomic For a perfect residuated lattice A, let A+ = (J(A), M(A), ≤, ◦, , , E) where x ◦ y = {z ∈ J(A) | z ≤ xy} and E = {z ∈ J(A) | z ≤ 1}

Theorem

A+ is a residuated frame and if A is complete then (A+)+ ∼ = A In particular, any finite lattice is complete and perfect So for finite residuated lattices, residuated frames give a compact representation analogous to atom structures for relation algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 8 / 39

slide-35
SLIDE 35

Frames of complete perfect lattices

A lattice L is perfect if every element is a join of elements of J(L) and a meet of elements of M(L) E.g. a Boolean algebra is perfect iff it is atomic For a perfect residuated lattice A, let A+ = (J(A), M(A), ≤, ◦, , , E) where x ◦ y = {z ∈ J(A) | z ≤ xy} and E = {z ∈ J(A) | z ≤ 1}

Theorem

A+ is a residuated frame and if A is complete then (A+)+ ∼ = A In particular, any finite lattice is complete and perfect So for finite residuated lattices, residuated frames give a compact representation analogous to atom structures for relation algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 8 / 39

slide-36
SLIDE 36

Frames of complete perfect lattices

A lattice L is perfect if every element is a join of elements of J(L) and a meet of elements of M(L) E.g. a Boolean algebra is perfect iff it is atomic For a perfect residuated lattice A, let A+ = (J(A), M(A), ≤, ◦, , , E) where x ◦ y = {z ∈ J(A) | z ≤ xy} and E = {z ∈ J(A) | z ≤ 1}

Theorem

A+ is a residuated frame and if A is complete then (A+)+ ∼ = A In particular, any finite lattice is complete and perfect So for finite residuated lattices, residuated frames give a compact representation analogous to atom structures for relation algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 8 / 39

slide-37
SLIDE 37

All (dually-)nonisomorphic lattices with ≤ 7 elements

K1,1 Kn,k = selfdual lattice of size n Ln,k = nonselfdual lattices of size n K2,1 K3,1 K4,1 K4,2 L5,1 K5,1 K5,2 K5,3 L6,1 L6,2 L6,3 L6,4 K6,1 K6,2 K6,3 K6,4 K6,5 K6,6 K6,7 L7,1 L7,2 L7,3 L7,4 L7,5 L7,6 L7,7 L7,8 L7,9 L7,10 L7,11 L7,12 L7,13 L7,14 L7,15 L7,16 L7,17 L7,18 L7,19 L7,20 K7,1 K7,2 K7,3 K7,4 K7,5 K7,6 K7,7 K7,8 K7,9 K7,10 K7,11 K7,12K7,13

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 9 / 39

slide-38
SLIDE 38

FL

x ⇒ a y◦a◦z ⇒ c y◦x◦z ⇒ c (cut) a ⇒ a (Id) y◦a◦z ⇒ c y◦a ∧ b◦z ⇒ c (∧Lℓ) y◦b◦z ⇒ c y◦a ∧ b◦z ⇒ c (∧Lr) x ⇒ a x ⇒ b x ⇒ a ∧ b (∧R) y◦a◦z ⇒ c y◦b◦z ⇒ c y◦a ∨ b◦z ⇒ c (∨L) x ⇒ a x ⇒ a ∨ b (∨Rℓ) x ⇒ b x ⇒ a ∨ b (∨Rr) x ⇒ a y◦b◦z ⇒ c y◦x ◦ (a\b)◦z ⇒ c (\L) a ◦ x ⇒ b x ⇒ a\b (\R) x ⇒ a y◦b◦z ⇒ c y◦(b/a) ◦ x◦z ⇒ c (/L) x ◦ a ⇒ b x ⇒ b/a (/R) y◦a ◦ b◦z ⇒ c y◦a · b◦z ⇒ c (·L) x ⇒ a y ⇒ b x ◦ y ⇒ a · b (·R) y ◦ z ⇒ a y◦1◦z ⇒ a (1L) ε ⇒ 1 (1R) where a, b, c ∈ Fm, x, y, z ∈ Fm∗.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 10 / 39

slide-39
SLIDE 39

FL with context notation

x ⇒ a u[a] ⇒ c u[x] ⇒ c (cut) a ⇒ a (Id) u[a] ⇒ c u[a ∧ b] ⇒ c (∧Lℓ) u[b] ⇒ c u[a ∧ b] ⇒ c (∧Lr) x ⇒ a x ⇒ b x ⇒ a ∧ b (∧R) u[a] ⇒ c u[b] ⇒ c u[a ∨ b] ⇒ c (∨L) x ⇒ a x ⇒ a ∨ b (∨Rℓ) x ⇒ b x ⇒ a ∨ b (∨Rr) x ⇒ a u[b] ⇒ c u[x ◦ (a\b)] ⇒ c (\L) a ◦ x ⇒ b x ⇒ a\b (\R) x ⇒ a u[b] ⇒ c u[(b/a) ◦ x] ⇒ c (/L) x ◦ a ⇒ b x ⇒ b/a (/R) u[a ◦ b] ⇒ c u[a · b] ⇒ c (·L) x ⇒ a y ⇒ b x ◦ y ⇒ a · b (·R) |u| ⇒ a u[1] ⇒ a (1L) ε ⇒ 1 (1R)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 11 / 39

slide-40
SLIDE 40

Basic substructural logics

If the sequent s is provable in FL from the set of sequents S, we write S ⊢FL s. u[x ◦ y] ⇒ c u[y ◦ x] ⇒ c (e) (exchange) xy ≤ yx u[x ◦ x] ⇒ c u[x] ⇒ c (c) (contraction) x ≤ x2 |u| ⇒ c u[x] ⇒ c (i) (integrality) x ≤ 1 We write FLec for FL + (e) + (c).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 12 / 39

slide-41
SLIDE 41

Basic substructural logics

If the sequent s is provable in FL from the set of sequents S, we write S ⊢FL s. u[x ◦ y] ⇒ c u[y ◦ x] ⇒ c (e) (exchange) xy ≤ yx u[x ◦ x] ⇒ c u[x] ⇒ c (c) (contraction) x ≤ x2 |u| ⇒ c u[x] ⇒ c (i) (integrality) x ≤ 1 We write FLec for FL + (e) + (c).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 12 / 39

slide-42
SLIDE 42

Basic substructural logics

If the sequent s is provable in FL from the set of sequents S, we write S ⊢FL s. u[x ◦ y] ⇒ c u[y ◦ x] ⇒ c (e) (exchange) xy ≤ yx u[x ◦ x] ⇒ c u[x] ⇒ c (c) (contraction) x ≤ x2 |u| ⇒ c u[x] ⇒ c (i) (integrality) x ≤ 1 We write FLec for FL + (e) + (c).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 12 / 39

slide-43
SLIDE 43

Basic substructural logics

If the sequent s is provable in FL from the set of sequents S, we write S ⊢FL s. u[x ◦ y] ⇒ c u[y ◦ x] ⇒ c (e) (exchange) xy ≤ yx u[x ◦ x] ⇒ c u[x] ⇒ c (c) (contraction) x ≤ x2 |u| ⇒ c u[x] ⇒ c (i) (integrality) x ≤ 1 We write FLec for FL + (e) + (c).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 12 / 39

slide-44
SLIDE 44

Basic substructural logics

If the sequent s is provable in FL from the set of sequents S, we write S ⊢FL s. u[x ◦ y] ⇒ c u[y ◦ x] ⇒ c (e) (exchange) xy ≤ yx u[x ◦ x] ⇒ c u[x] ⇒ c (c) (contraction) x ≤ x2 |u| ⇒ c u[x] ⇒ c (i) (integrality) x ≤ 1 We write FLec for FL + (e) + (c).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 12 / 39

slide-45
SLIDE 45

Examples of frames (FL)

Consider the Gentzen system FL (full Lambek calculus). We define the frame WFL, where (W , ◦, ε) to be the free monoid over the set Fm of all formulas W ′ = SW × Fm, where SW is the set of all unary linear polynomials u[x] = y◦x◦z of W , and x N (u, a) iff ⊢FL u[x] ⇒ a. For (u, a) x = {(u[ ◦ x], a)} and x (u, a) = {(u[x ◦ ], a)}, we have x ◦ yN(u, a) iff ⊢FL u[x ◦ y] ⇒ a iff ⊢FL u[x◦y] ⇒ a iff xN(u[ ◦ y], a) iff yN(u[x ◦ ], a).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 13 / 39

slide-46
SLIDE 46

Examples of frames (FL)

Consider the Gentzen system FL (full Lambek calculus). We define the frame WFL, where (W , ◦, ε) to be the free monoid over the set Fm of all formulas W ′ = SW × Fm, where SW is the set of all unary linear polynomials u[x] = y◦x◦z of W , and x N (u, a) iff ⊢FL u[x] ⇒ a. For (u, a) x = {(u[ ◦ x], a)} and x (u, a) = {(u[x ◦ ], a)}, we have x ◦ yN(u, a) iff ⊢FL u[x ◦ y] ⇒ a iff ⊢FL u[x◦y] ⇒ a iff xN(u[ ◦ y], a) iff yN(u[x ◦ ], a).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 13 / 39

slide-47
SLIDE 47

Examples of frames (FL)

Consider the Gentzen system FL (full Lambek calculus). We define the frame WFL, where (W , ◦, ε) to be the free monoid over the set Fm of all formulas W ′ = SW × Fm, where SW is the set of all unary linear polynomials u[x] = y◦x◦z of W , and x N (u, a) iff ⊢FL u[x] ⇒ a. For (u, a) x = {(u[ ◦ x], a)} and x (u, a) = {(u[x ◦ ], a)}, we have x ◦ yN(u, a) iff ⊢FL u[x ◦ y] ⇒ a iff ⊢FL u[x◦y] ⇒ a iff xN(u[ ◦ y], a) iff yN(u[x ◦ ], a).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 13 / 39

slide-48
SLIDE 48

Examples of frames (FL)

Consider the Gentzen system FL (full Lambek calculus). We define the frame WFL, where (W , ◦, ε) to be the free monoid over the set Fm of all formulas W ′ = SW × Fm, where SW is the set of all unary linear polynomials u[x] = y◦x◦z of W , and x N (u, a) iff ⊢FL u[x] ⇒ a. For (u, a) x = {(u[ ◦ x], a)} and x (u, a) = {(u[x ◦ ], a)}, we have x ◦ yN(u, a) iff ⊢FL u[x ◦ y] ⇒ a iff ⊢FL u[x◦y] ⇒ a iff xN(u[ ◦ y], a) iff yN(u[x ◦ ], a).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 13 / 39

slide-49
SLIDE 49

Examples of frames (FL)

Consider the Gentzen system FL (full Lambek calculus). We define the frame WFL, where (W , ◦, ε) to be the free monoid over the set Fm of all formulas W ′ = SW × Fm, where SW is the set of all unary linear polynomials u[x] = y◦x◦z of W , and x N (u, a) iff ⊢FL u[x] ⇒ a. For (u, a) x = {(u[ ◦ x], a)} and x (u, a) = {(u[x ◦ ], a)}, we have x ◦ yN(u, a) iff ⊢FL u[x ◦ y] ⇒ a iff ⊢FL u[x◦y] ⇒ a iff xN(u[ ◦ y], a) iff yN(u[x ◦ ], a).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 13 / 39

slide-50
SLIDE 50

Examples of frames (FEP)

Let A be a residuated lattice and B a partial subalgebra of A. We define the frame WA,B, where (W , ·, 1) to be the submonoid of A generated by B, W ′ = SB × B, where SW is the set of all unary linear polynomials u[x] = y◦x◦z of (W , ·, 1), and x N (u, b) by u[x] ≤A b. For (u, a) x = {(u[ · x], a)} and x (u, a) = {(u[x · ], a)}, we have x · yN(u, a) iff u[x · y] ≤ a iff xN(u[ · y], a) iff yN(u[x · ], a).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 14 / 39

slide-51
SLIDE 51

Examples of frames (FEP)

Let A be a residuated lattice and B a partial subalgebra of A. We define the frame WA,B, where (W , ·, 1) to be the submonoid of A generated by B, W ′ = SB × B, where SW is the set of all unary linear polynomials u[x] = y◦x◦z of (W , ·, 1), and x N (u, b) by u[x] ≤A b. For (u, a) x = {(u[ · x], a)} and x (u, a) = {(u[x · ], a)}, we have x · yN(u, a) iff u[x · y] ≤ a iff xN(u[ · y], a) iff yN(u[x · ], a).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 14 / 39

slide-52
SLIDE 52

Examples of frames (FEP)

Let A be a residuated lattice and B a partial subalgebra of A. We define the frame WA,B, where (W , ·, 1) to be the submonoid of A generated by B, W ′ = SB × B, where SW is the set of all unary linear polynomials u[x] = y◦x◦z of (W , ·, 1), and x N (u, b) by u[x] ≤A b. For (u, a) x = {(u[ · x], a)} and x (u, a) = {(u[x · ], a)}, we have x · yN(u, a) iff u[x · y] ≤ a iff xN(u[ · y], a) iff yN(u[x · ], a).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 14 / 39

slide-53
SLIDE 53

Examples of frames (FEP)

Let A be a residuated lattice and B a partial subalgebra of A. We define the frame WA,B, where (W , ·, 1) to be the submonoid of A generated by B, W ′ = SB × B, where SW is the set of all unary linear polynomials u[x] = y◦x◦z of (W , ·, 1), and x N (u, b) by u[x] ≤A b. For (u, a) x = {(u[ · x], a)} and x (u, a) = {(u[x · ], a)}, we have x · yN(u, a) iff u[x · y] ≤ a iff xN(u[ · y], a) iff yN(u[x · ], a).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 14 / 39

slide-54
SLIDE 54

Examples of frames (FEP)

Let A be a residuated lattice and B a partial subalgebra of A. We define the frame WA,B, where (W , ·, 1) to be the submonoid of A generated by B, W ′ = SB × B, where SW is the set of all unary linear polynomials u[x] = y◦x◦z of (W , ·, 1), and x N (u, b) by u[x] ≤A b. For (u, a) x = {(u[ · x], a)} and x (u, a) = {(u[x · ], a)}, we have x · yN(u, a) iff u[x · y] ≤ a iff xN(u[ · y], a) iff yN(u[x · ], a).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 14 / 39

slide-55
SLIDE 55

GN

xNa aNz xNz (CUT) aNa (Id) xNa bNz x ◦ (a\b)Nz (\L) a ◦ xNb xNa\b (\R) xNa bNz (b/a) ◦ xNz (/L) x ◦ aNb xNb/a (/R) a ◦ bNz a · bNz (·L) xNa yNb x ◦ yNa · b (·R) aNz a ∧ bNz (∧Lℓ) bNz a ∧ bNz (∧Lr) xNa xNb xNa ∧ b (∧R) aNz bNz a ∨ bNz (∨L) xNa xNa ∨ b (∨Rℓ) xNb xNa ∨ b (∨Rr) εNz 1Nz (1L) εN1 (1R)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 15 / 39

slide-56
SLIDE 56

Gentzen frames

The following properties hold for WL, WFL and WA,B:

1 W is a residuated frame 2 B is a (partial) algebra of the same type, (B = L, Fm, B) 3 B generates (W , ◦, ε) (as a monoid) 4 W ′ contains a copy of B (b ↔ (id, b)) 5 N satisfies GN, for all a, b ∈ B, x, y ∈ W , z ∈ W ′.

We call such pairs (W, B) Gentzen frames. A cut-free Gentzen frame is not assumed to satisfy the (CUT)-rule.

  • Theorem. Given a Gentzen frame (W, B), the map

{}⊳ : B → W+, b → {b}⊳ is a (partial) homomorphism. (Namely, if a, b ∈ B and a • b ∈ B (• is a connective) then {a •B b}⊳ = {a}⊳ •W+ {b}⊳).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 16 / 39

slide-57
SLIDE 57

Gentzen frames

The following properties hold for WL, WFL and WA,B:

1 W is a residuated frame 2 B is a (partial) algebra of the same type, (B = L, Fm, B) 3 B generates (W , ◦, ε) (as a monoid) 4 W ′ contains a copy of B (b ↔ (id, b)) 5 N satisfies GN, for all a, b ∈ B, x, y ∈ W , z ∈ W ′.

We call such pairs (W, B) Gentzen frames. A cut-free Gentzen frame is not assumed to satisfy the (CUT)-rule.

  • Theorem. Given a Gentzen frame (W, B), the map

{}⊳ : B → W+, b → {b}⊳ is a (partial) homomorphism. (Namely, if a, b ∈ B and a • b ∈ B (• is a connective) then {a •B b}⊳ = {a}⊳ •W+ {b}⊳).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 16 / 39

slide-58
SLIDE 58

Gentzen frames

The following properties hold for WL, WFL and WA,B:

1 W is a residuated frame 2 B is a (partial) algebra of the same type, (B = L, Fm, B) 3 B generates (W , ◦, ε) (as a monoid) 4 W ′ contains a copy of B (b ↔ (id, b)) 5 N satisfies GN, for all a, b ∈ B, x, y ∈ W , z ∈ W ′.

We call such pairs (W, B) Gentzen frames. A cut-free Gentzen frame is not assumed to satisfy the (CUT)-rule.

  • Theorem. Given a Gentzen frame (W, B), the map

{}⊳ : B → W+, b → {b}⊳ is a (partial) homomorphism. (Namely, if a, b ∈ B and a • b ∈ B (• is a connective) then {a •B b}⊳ = {a}⊳ •W+ {b}⊳).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 16 / 39

slide-59
SLIDE 59

Proof

Key Lemma. Let (W, B) be a Gentzen frame. For all a, b ∈ B, k, l ∈ W+ and for every connective •, if a • b ∈ B, a ∈ X ⊆ {a}⊳ and b ∈ Y ⊆ {b}⊳, then

1 a •B b ∈ X •W+ Y ⊆ {a •B b}⊳ (1B ∈ 1W+ ⊆ {1B}⊳ ) 2 In particular, a •B b ∈ {a}⊳ •W+ {b}⊳ ⊆ {a •B b}⊳. 3 Furthermore, because of (CUT), we have equality.

Proof Let • = ∨. If x ∈ X, then x ∈ {a}⊳; so xNa and xNa ∨ b, by (∨Rℓ); hence x ∈ {a ∨ b}⊳ and X ⊆ {a ∨ b}⊳. Likewise Y ⊆ {a ∨ b}⊳, so X ∪ Y ⊆ {a ∨ b}⊳ and X ∨ Y = γ(X ∪ Y ) ⊆ {a ∨ b}⊳. On the other hand, let X ∨ Y ⊆ {z}⊳, for some z ∈ W . Then, a ∈ X ⊆ X ∨ Y ⊆ {z}⊳, so aNz. Similarly, bNz, so a ∨ bNz by (∨L), hence a ∨ b ∈ {z}⊳. Thus, a ∨ b ∈ X ∨ Y . We used that every closed set is an intersection of basic closed sets {z}⊳, for z ∈ W .

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 17 / 39

slide-60
SLIDE 60

Proof

Key Lemma. Let (W, B) be a Gentzen frame. For all a, b ∈ B, k, l ∈ W+ and for every connective •, if a • b ∈ B, a ∈ X ⊆ {a}⊳ and b ∈ Y ⊆ {b}⊳, then

1 a •B b ∈ X •W+ Y ⊆ {a •B b}⊳ (1B ∈ 1W+ ⊆ {1B}⊳ ) 2 In particular, a •B b ∈ {a}⊳ •W+ {b}⊳ ⊆ {a •B b}⊳. 3 Furthermore, because of (CUT), we have equality.

Proof Let • = ∨. If x ∈ X, then x ∈ {a}⊳; so xNa and xNa ∨ b, by (∨Rℓ); hence x ∈ {a ∨ b}⊳ and X ⊆ {a ∨ b}⊳. Likewise Y ⊆ {a ∨ b}⊳, so X ∪ Y ⊆ {a ∨ b}⊳ and X ∨ Y = γ(X ∪ Y ) ⊆ {a ∨ b}⊳. On the other hand, let X ∨ Y ⊆ {z}⊳, for some z ∈ W . Then, a ∈ X ⊆ X ∨ Y ⊆ {z}⊳, so aNz. Similarly, bNz, so a ∨ bNz by (∨L), hence a ∨ b ∈ {z}⊳. Thus, a ∨ b ∈ X ∨ Y . We used that every closed set is an intersection of basic closed sets {z}⊳, for z ∈ W .

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 17 / 39

slide-61
SLIDE 61

Proof

Key Lemma. Let (W, B) be a Gentzen frame. For all a, b ∈ B, k, l ∈ W+ and for every connective •, if a • b ∈ B, a ∈ X ⊆ {a}⊳ and b ∈ Y ⊆ {b}⊳, then

1 a •B b ∈ X •W+ Y ⊆ {a •B b}⊳ (1B ∈ 1W+ ⊆ {1B}⊳ ) 2 In particular, a •B b ∈ {a}⊳ •W+ {b}⊳ ⊆ {a •B b}⊳. 3 Furthermore, because of (CUT), we have equality.

Proof Let • = ∨. If x ∈ X, then x ∈ {a}⊳; so xNa and xNa ∨ b, by (∨Rℓ); hence x ∈ {a ∨ b}⊳ and X ⊆ {a ∨ b}⊳. Likewise Y ⊆ {a ∨ b}⊳, so X ∪ Y ⊆ {a ∨ b}⊳ and X ∨ Y = γ(X ∪ Y ) ⊆ {a ∨ b}⊳. On the other hand, let X ∨ Y ⊆ {z}⊳, for some z ∈ W . Then, a ∈ X ⊆ X ∨ Y ⊆ {z}⊳, so aNz. Similarly, bNz, so a ∨ bNz by (∨L), hence a ∨ b ∈ {z}⊳. Thus, a ∨ b ∈ X ∨ Y . We used that every closed set is an intersection of basic closed sets {z}⊳, for z ∈ W .

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 17 / 39

slide-62
SLIDE 62

DM-completion

For a residuated lattice L, we associated the Gentzen frame (WL, L). The underlying poset of W+

L is the Dedekind-MacNeille completion of the

underlying poset reduct of L.

  • Theorem. The map x → x⊳ is an embedding of L into W+

L .

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 18 / 39

slide-63
SLIDE 63

DM-completion

For a residuated lattice L, we associated the Gentzen frame (WL, L). The underlying poset of W+

L is the Dedekind-MacNeille completion of the

underlying poset reduct of L.

  • Theorem. The map x → x⊳ is an embedding of L into W+

L .

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 18 / 39

slide-64
SLIDE 64

Completeness - Cut elimination

For every homomorphism f : Fm → B, let ¯ f : FmL → W+ be the homomorphism that extends ¯ f (p) = {f (p)}⊳ for any variable p

  • Corollary. If (W, B) is a cf Gentzen frame then for every homomorphism

f : Fm → B, we have f (a) ∈ ¯ f (a) ⊆ {f (a)}⊳. With CUT ¯ f (a)={f (a)}⊳ We define WFL | = x ⇒ c if f (x) N f (c) for all f : Fm → Fm

  • Theorem. If W+

FL |

= x· ≤ c, then WFL | = x ⇒ c. Idea: For f : Fm → B, f (x) ∈ ¯ f (x) ⊆ ¯ f (c) ⊆ {f (c)}⊳, so f (x) N f (c).

  • Corollary. FL is complete with respect to W+

FL.

  • Corollary. The algebra W+

FL generates RL.

The frame WFLf corresponds to cut-free FL. Corollary (CE). FL and FLf prove the same sequents.

  • Corollary. FL and the equational theory of RL are decidable.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 19 / 39

slide-65
SLIDE 65

Completeness - Cut elimination

For every homomorphism f : Fm → B, let ¯ f : FmL → W+ be the homomorphism that extends ¯ f (p) = {f (p)}⊳ for any variable p

  • Corollary. If (W, B) is a cf Gentzen frame then for every homomorphism

f : Fm → B, we have f (a) ∈ ¯ f (a) ⊆ {f (a)}⊳. With CUT ¯ f (a)={f (a)}⊳ We define WFL | = x ⇒ c if f (x) N f (c) for all f : Fm → Fm

  • Theorem. If W+

FL |

= x· ≤ c, then WFL | = x ⇒ c. Idea: For f : Fm → B, f (x) ∈ ¯ f (x) ⊆ ¯ f (c) ⊆ {f (c)}⊳, so f (x) N f (c).

  • Corollary. FL is complete with respect to W+

FL.

  • Corollary. The algebra W+

FL generates RL.

The frame WFLf corresponds to cut-free FL. Corollary (CE). FL and FLf prove the same sequents.

  • Corollary. FL and the equational theory of RL are decidable.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 19 / 39

slide-66
SLIDE 66

Completeness - Cut elimination

For every homomorphism f : Fm → B, let ¯ f : FmL → W+ be the homomorphism that extends ¯ f (p) = {f (p)}⊳ for any variable p

  • Corollary. If (W, B) is a cf Gentzen frame then for every homomorphism

f : Fm → B, we have f (a) ∈ ¯ f (a) ⊆ {f (a)}⊳. With CUT ¯ f (a)={f (a)}⊳ We define WFL | = x ⇒ c if f (x) N f (c) for all f : Fm → Fm

  • Theorem. If W+

FL |

= x· ≤ c, then WFL | = x ⇒ c. Idea: For f : Fm → B, f (x) ∈ ¯ f (x) ⊆ ¯ f (c) ⊆ {f (c)}⊳, so f (x) N f (c).

  • Corollary. FL is complete with respect to W+

FL.

  • Corollary. The algebra W+

FL generates RL.

The frame WFLf corresponds to cut-free FL. Corollary (CE). FL and FLf prove the same sequents.

  • Corollary. FL and the equational theory of RL are decidable.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 19 / 39

slide-67
SLIDE 67

Completeness - Cut elimination

For every homomorphism f : Fm → B, let ¯ f : FmL → W+ be the homomorphism that extends ¯ f (p) = {f (p)}⊳ for any variable p

  • Corollary. If (W, B) is a cf Gentzen frame then for every homomorphism

f : Fm → B, we have f (a) ∈ ¯ f (a) ⊆ {f (a)}⊳. With CUT ¯ f (a)={f (a)}⊳ We define WFL | = x ⇒ c if f (x) N f (c) for all f : Fm → Fm

  • Theorem. If W+

FL |

= x· ≤ c, then WFL | = x ⇒ c. Idea: For f : Fm → B, f (x) ∈ ¯ f (x) ⊆ ¯ f (c) ⊆ {f (c)}⊳, so f (x) N f (c).

  • Corollary. FL is complete with respect to W+

FL.

  • Corollary. The algebra W+

FL generates RL.

The frame WFLf corresponds to cut-free FL. Corollary (CE). FL and FLf prove the same sequents.

  • Corollary. FL and the equational theory of RL are decidable.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 19 / 39

slide-68
SLIDE 68

Completeness - Cut elimination

For every homomorphism f : Fm → B, let ¯ f : FmL → W+ be the homomorphism that extends ¯ f (p) = {f (p)}⊳ for any variable p

  • Corollary. If (W, B) is a cf Gentzen frame then for every homomorphism

f : Fm → B, we have f (a) ∈ ¯ f (a) ⊆ {f (a)}⊳. With CUT ¯ f (a)={f (a)}⊳ We define WFL | = x ⇒ c if f (x) N f (c) for all f : Fm → Fm

  • Theorem. If W+

FL |

= x· ≤ c, then WFL | = x ⇒ c. Idea: For f : Fm → B, f (x) ∈ ¯ f (x) ⊆ ¯ f (c) ⊆ {f (c)}⊳, so f (x) N f (c).

  • Corollary. FL is complete with respect to W+

FL.

  • Corollary. The algebra W+

FL generates RL.

The frame WFLf corresponds to cut-free FL. Corollary (CE). FL and FLf prove the same sequents.

  • Corollary. FL and the equational theory of RL are decidable.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 19 / 39

slide-69
SLIDE 69

Finite model property

For WFL, given (x, z) ∈ W × W ′ (if z = (u, c), then u(x) ⇒ c is a sequent), we define (x, z)↑ as the smallest subset of W × W ′ that contains (x, z) and is closed upwards with respect to the rules of FLf. Note that (x, z)↑ is finite. The new frame W′ associated with N′ = N ∪ ((y, v)↑)c is residuated and Gentzen. (N′)c is finite, so has finite domain Dom((N′)c) and codomain Cod((N′)c) For every z ∈ Cod((N′)c), {z}⊳ = W . So, {{z}⊳ : z ∈ W } is finite and a basis for γN′. So W′+ is finite. Moreover, if u(x) ⇒ c is not provable in FL, then it is not valid in W′+.

  • Corollary. The system FL has the finite model property.
  • Corollary. The variety RL is generated by its finite members.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 20 / 39

slide-70
SLIDE 70

Finite model property

For WFL, given (x, z) ∈ W × W ′ (if z = (u, c), then u(x) ⇒ c is a sequent), we define (x, z)↑ as the smallest subset of W × W ′ that contains (x, z) and is closed upwards with respect to the rules of FLf. Note that (x, z)↑ is finite. The new frame W′ associated with N′ = N ∪ ((y, v)↑)c is residuated and Gentzen. (N′)c is finite, so has finite domain Dom((N′)c) and codomain Cod((N′)c) For every z ∈ Cod((N′)c), {z}⊳ = W . So, {{z}⊳ : z ∈ W } is finite and a basis for γN′. So W′+ is finite. Moreover, if u(x) ⇒ c is not provable in FL, then it is not valid in W′+.

  • Corollary. The system FL has the finite model property.
  • Corollary. The variety RL is generated by its finite members.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 20 / 39

slide-71
SLIDE 71

Finite model property

For WFL, given (x, z) ∈ W × W ′ (if z = (u, c), then u(x) ⇒ c is a sequent), we define (x, z)↑ as the smallest subset of W × W ′ that contains (x, z) and is closed upwards with respect to the rules of FLf. Note that (x, z)↑ is finite. The new frame W′ associated with N′ = N ∪ ((y, v)↑)c is residuated and Gentzen. (N′)c is finite, so has finite domain Dom((N′)c) and codomain Cod((N′)c) For every z ∈ Cod((N′)c), {z}⊳ = W . So, {{z}⊳ : z ∈ W } is finite and a basis for γN′. So W′+ is finite. Moreover, if u(x) ⇒ c is not provable in FL, then it is not valid in W′+.

  • Corollary. The system FL has the finite model property.
  • Corollary. The variety RL is generated by its finite members.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 20 / 39

slide-72
SLIDE 72

Finite model property

For WFL, given (x, z) ∈ W × W ′ (if z = (u, c), then u(x) ⇒ c is a sequent), we define (x, z)↑ as the smallest subset of W × W ′ that contains (x, z) and is closed upwards with respect to the rules of FLf. Note that (x, z)↑ is finite. The new frame W′ associated with N′ = N ∪ ((y, v)↑)c is residuated and Gentzen. (N′)c is finite, so has finite domain Dom((N′)c) and codomain Cod((N′)c) For every z ∈ Cod((N′)c), {z}⊳ = W . So, {{z}⊳ : z ∈ W } is finite and a basis for γN′. So W′+ is finite. Moreover, if u(x) ⇒ c is not provable in FL, then it is not valid in W′+.

  • Corollary. The system FL has the finite model property.
  • Corollary. The variety RL is generated by its finite members.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 20 / 39

slide-73
SLIDE 73

FEP

A class of algebras K has the finite embeddability property (FEP) if for every A ∈ K, every finite partial subalgebra B of A can be (partially) embedded in a finite D ∈ K. The corresponding logic has the strong finite model property: if Φ ⊢ ψ, for finite Φ, then there is a finite counter-model, namely there is D ∈ K and a homomorphism f : Fm → D, such that f (φ) = 1, for all φ ∈ Φ, but f (ψ) = 1. For logics with finitely many axioms and rules, this implies that the deducibility relation is decidable For a finitely axiomatized variety, FEP implies the decidability of the universal theory

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 21 / 39

slide-74
SLIDE 74

FEP

A class of algebras K has the finite embeddability property (FEP) if for every A ∈ K, every finite partial subalgebra B of A can be (partially) embedded in a finite D ∈ K. The corresponding logic has the strong finite model property: if Φ ⊢ ψ, for finite Φ, then there is a finite counter-model, namely there is D ∈ K and a homomorphism f : Fm → D, such that f (φ) = 1, for all φ ∈ Φ, but f (ψ) = 1. For logics with finitely many axioms and rules, this implies that the deducibility relation is decidable For a finitely axiomatized variety, FEP implies the decidability of the universal theory

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 21 / 39

slide-75
SLIDE 75

FEP

A class of algebras K has the finite embeddability property (FEP) if for every A ∈ K, every finite partial subalgebra B of A can be (partially) embedded in a finite D ∈ K. The corresponding logic has the strong finite model property: if Φ ⊢ ψ, for finite Φ, then there is a finite counter-model, namely there is D ∈ K and a homomorphism f : Fm → D, such that f (φ) = 1, for all φ ∈ Φ, but f (ψ) = 1. For logics with finitely many axioms and rules, this implies that the deducibility relation is decidable For a finitely axiomatized variety, FEP implies the decidability of the universal theory

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 21 / 39

slide-76
SLIDE 76

FEP

A class of algebras K has the finite embeddability property (FEP) if for every A ∈ K, every finite partial subalgebra B of A can be (partially) embedded in a finite D ∈ K. The corresponding logic has the strong finite model property: if Φ ⊢ ψ, for finite Φ, then there is a finite counter-model, namely there is D ∈ K and a homomorphism f : Fm → D, such that f (φ) = 1, for all φ ∈ Φ, but f (ψ) = 1. For logics with finitely many axioms and rules, this implies that the deducibility relation is decidable For a finitely axiomatized variety, FEP implies the decidability of the universal theory

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 21 / 39

slide-77
SLIDE 77

FEP for integral RLs with {∨, ·, 1}-equations

Blok and van Alten 2002 proved FEP for integral RLs, and extended it to residuated groupoids (2005)

  • Theorem. Every variety of integral RL’s axiomatized by equations over

{∨, ·, 1} has the FEP. B embeds in W+

A,B via { }⊳ : B → W+

W+

A,B is finite

W+

A,B ∈ V

  • Corollary. These varieties are generated as quasivarieties by their finite

members.

  • Corollary. The corresponding logics have the strong finite model property

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 22 / 39

slide-78
SLIDE 78

FEP for integral RLs with {∨, ·, 1}-equations

Blok and van Alten 2002 proved FEP for integral RLs, and extended it to residuated groupoids (2005)

  • Theorem. Every variety of integral RL’s axiomatized by equations over

{∨, ·, 1} has the FEP. B embeds in W+

A,B via { }⊳ : B → W+

W+

A,B is finite

W+

A,B ∈ V

  • Corollary. These varieties are generated as quasivarieties by their finite

members.

  • Corollary. The corresponding logics have the strong finite model property

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 22 / 39

slide-79
SLIDE 79

FEP for integral RLs with {∨, ·, 1}-equations

Blok and van Alten 2002 proved FEP for integral RLs, and extended it to residuated groupoids (2005)

  • Theorem. Every variety of integral RL’s axiomatized by equations over

{∨, ·, 1} has the FEP. B embeds in W+

A,B via { }⊳ : B → W+

W+

A,B is finite

W+

A,B ∈ V

  • Corollary. These varieties are generated as quasivarieties by their finite

members.

  • Corollary. The corresponding logics have the strong finite model property

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 22 / 39

slide-80
SLIDE 80

Finiteness

Idea for finiteness: Every element in W+

A,B is an intersection of basic

  • elements. So it suffices to prove that there are only finitely many such

elements. Replace the frame WA,B by one WM

A,B, where it is easier to work.

Let M be the free monoid with unit over the set B and f : M → W the extension of the identity map. M

f

− → W

N

− − W ′ .

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 23 / 39

slide-81
SLIDE 81

Finiteness

Idea for finiteness: Every element in W+

A,B is an intersection of basic

  • elements. So it suffices to prove that there are only finitely many such

elements. Replace the frame WA,B by one WM

A,B, where it is easier to work.

Let M be the free monoid with unit over the set B and f : M → W the extension of the identity map. M

f

− → W

N

− − W ′ .

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 23 / 39

slide-82
SLIDE 82

Finiteness

Idea for finiteness: Every element in W+

A,B is an intersection of basic

  • elements. So it suffices to prove that there are only finitely many such

elements. Replace the frame WA,B by one WM

A,B, where it is easier to work.

Let M be the free monoid with unit over the set B and f : M → W the extension of the identity map. M

f

− → W

N

− − W ′ .

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 23 / 39

slide-83
SLIDE 83

Equations 1

Idea: Express equations over {∨, ·, 1} at the frame level. For an equation ε over {∨, ·, 1} we distribute products over joins to get s1 ∨ · · · ∨ sm = t1 ∨ · · · ∨ tn. si, tj: monoid terms. s1 ∨ · · · ∨ sm ≤ t1 ∨ · · · ∨ tn and t1 ∨ · · · ∨ tn ≤ s1 ∨ · · · ∨ sm. The first is equivalent to: &(sj ≤ t1 ∨ · · · ∨ tn). We proceed by example: x2y ≤ xy ∨ yx (x1 ∨ x2)2y ≤ (x1 ∨ x2)y ∨ y(x1 ∨ x2) x2

1y ∨ x1x2y ∨ x2x1y ∨ x2 2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2

x1x2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2 x1y ≤ v x2y ≤ v yx1 ≤ v yx2 ≤ v x1x2y ≤ v x1 ◦ y N z x2 ◦ y N z y ◦ x1 N z y ◦ x2 N z x1 ◦ x2 ◦ y N z R(ε)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 24 / 39

slide-84
SLIDE 84

Equations 1

Idea: Express equations over {∨, ·, 1} at the frame level. For an equation ε over {∨, ·, 1} we distribute products over joins to get s1 ∨ · · · ∨ sm = t1 ∨ · · · ∨ tn. si, tj: monoid terms. s1 ∨ · · · ∨ sm ≤ t1 ∨ · · · ∨ tn and t1 ∨ · · · ∨ tn ≤ s1 ∨ · · · ∨ sm. The first is equivalent to: &(sj ≤ t1 ∨ · · · ∨ tn). We proceed by example: x2y ≤ xy ∨ yx (x1 ∨ x2)2y ≤ (x1 ∨ x2)y ∨ y(x1 ∨ x2) x2

1y ∨ x1x2y ∨ x2x1y ∨ x2 2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2

x1x2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2 x1y ≤ v x2y ≤ v yx1 ≤ v yx2 ≤ v x1x2y ≤ v x1 ◦ y N z x2 ◦ y N z y ◦ x1 N z y ◦ x2 N z x1 ◦ x2 ◦ y N z R(ε)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 24 / 39

slide-85
SLIDE 85

Equations 1

Idea: Express equations over {∨, ·, 1} at the frame level. For an equation ε over {∨, ·, 1} we distribute products over joins to get s1 ∨ · · · ∨ sm = t1 ∨ · · · ∨ tn. si, tj: monoid terms. s1 ∨ · · · ∨ sm ≤ t1 ∨ · · · ∨ tn and t1 ∨ · · · ∨ tn ≤ s1 ∨ · · · ∨ sm. The first is equivalent to: &(sj ≤ t1 ∨ · · · ∨ tn). We proceed by example: x2y ≤ xy ∨ yx (x1 ∨ x2)2y ≤ (x1 ∨ x2)y ∨ y(x1 ∨ x2) x2

1y ∨ x1x2y ∨ x2x1y ∨ x2 2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2

x1x2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2 x1y ≤ v x2y ≤ v yx1 ≤ v yx2 ≤ v x1x2y ≤ v x1 ◦ y N z x2 ◦ y N z y ◦ x1 N z y ◦ x2 N z x1 ◦ x2 ◦ y N z R(ε)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 24 / 39

slide-86
SLIDE 86

Equations 1

Idea: Express equations over {∨, ·, 1} at the frame level. For an equation ε over {∨, ·, 1} we distribute products over joins to get s1 ∨ · · · ∨ sm = t1 ∨ · · · ∨ tn. si, tj: monoid terms. s1 ∨ · · · ∨ sm ≤ t1 ∨ · · · ∨ tn and t1 ∨ · · · ∨ tn ≤ s1 ∨ · · · ∨ sm. The first is equivalent to: &(sj ≤ t1 ∨ · · · ∨ tn). We proceed by example: x2y ≤ xy ∨ yx (x1 ∨ x2)2y ≤ (x1 ∨ x2)y ∨ y(x1 ∨ x2) x2

1y ∨ x1x2y ∨ x2x1y ∨ x2 2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2

x1x2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2 x1y ≤ v x2y ≤ v yx1 ≤ v yx2 ≤ v x1x2y ≤ v x1 ◦ y N z x2 ◦ y N z y ◦ x1 N z y ◦ x2 N z x1 ◦ x2 ◦ y N z R(ε)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 24 / 39

slide-87
SLIDE 87

Equations 1

Idea: Express equations over {∨, ·, 1} at the frame level. For an equation ε over {∨, ·, 1} we distribute products over joins to get s1 ∨ · · · ∨ sm = t1 ∨ · · · ∨ tn. si, tj: monoid terms. s1 ∨ · · · ∨ sm ≤ t1 ∨ · · · ∨ tn and t1 ∨ · · · ∨ tn ≤ s1 ∨ · · · ∨ sm. The first is equivalent to: &(sj ≤ t1 ∨ · · · ∨ tn). We proceed by example: x2y ≤ xy ∨ yx (x1 ∨ x2)2y ≤ (x1 ∨ x2)y ∨ y(x1 ∨ x2) x2

1y ∨ x1x2y ∨ x2x1y ∨ x2 2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2

x1x2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2 x1y ≤ v x2y ≤ v yx1 ≤ v yx2 ≤ v x1x2y ≤ v x1 ◦ y N z x2 ◦ y N z y ◦ x1 N z y ◦ x2 N z x1 ◦ x2 ◦ y N z R(ε)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 24 / 39

slide-88
SLIDE 88

Equations 1

Idea: Express equations over {∨, ·, 1} at the frame level. For an equation ε over {∨, ·, 1} we distribute products over joins to get s1 ∨ · · · ∨ sm = t1 ∨ · · · ∨ tn. si, tj: monoid terms. s1 ∨ · · · ∨ sm ≤ t1 ∨ · · · ∨ tn and t1 ∨ · · · ∨ tn ≤ s1 ∨ · · · ∨ sm. The first is equivalent to: &(sj ≤ t1 ∨ · · · ∨ tn). We proceed by example: x2y ≤ xy ∨ yx (x1 ∨ x2)2y ≤ (x1 ∨ x2)y ∨ y(x1 ∨ x2) x2

1y ∨ x1x2y ∨ x2x1y ∨ x2 2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2

x1x2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2 x1y ≤ v x2y ≤ v yx1 ≤ v yx2 ≤ v x1x2y ≤ v x1 ◦ y N z x2 ◦ y N z y ◦ x1 N z y ◦ x2 N z x1 ◦ x2 ◦ y N z R(ε)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 24 / 39

slide-89
SLIDE 89

Equations 1

Idea: Express equations over {∨, ·, 1} at the frame level. For an equation ε over {∨, ·, 1} we distribute products over joins to get s1 ∨ · · · ∨ sm = t1 ∨ · · · ∨ tn. si, tj: monoid terms. s1 ∨ · · · ∨ sm ≤ t1 ∨ · · · ∨ tn and t1 ∨ · · · ∨ tn ≤ s1 ∨ · · · ∨ sm. The first is equivalent to: &(sj ≤ t1 ∨ · · · ∨ tn). We proceed by example: x2y ≤ xy ∨ yx (x1 ∨ x2)2y ≤ (x1 ∨ x2)y ∨ y(x1 ∨ x2) x2

1y ∨ x1x2y ∨ x2x1y ∨ x2 2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2

x1x2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2 x1y ≤ v x2y ≤ v yx1 ≤ v yx2 ≤ v x1x2y ≤ v x1 ◦ y N z x2 ◦ y N z y ◦ x1 N z y ◦ x2 N z x1 ◦ x2 ◦ y N z R(ε)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 24 / 39

slide-90
SLIDE 90

Equations 1

Idea: Express equations over {∨, ·, 1} at the frame level. For an equation ε over {∨, ·, 1} we distribute products over joins to get s1 ∨ · · · ∨ sm = t1 ∨ · · · ∨ tn. si, tj: monoid terms. s1 ∨ · · · ∨ sm ≤ t1 ∨ · · · ∨ tn and t1 ∨ · · · ∨ tn ≤ s1 ∨ · · · ∨ sm. The first is equivalent to: &(sj ≤ t1 ∨ · · · ∨ tn). We proceed by example: x2y ≤ xy ∨ yx (x1 ∨ x2)2y ≤ (x1 ∨ x2)y ∨ y(x1 ∨ x2) x2

1y ∨ x1x2y ∨ x2x1y ∨ x2 2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2

x1x2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2 x1y ≤ v x2y ≤ v yx1 ≤ v yx2 ≤ v x1x2y ≤ v x1 ◦ y N z x2 ◦ y N z y ◦ x1 N z y ◦ x2 N z x1 ◦ x2 ◦ y N z R(ε)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 24 / 39

slide-91
SLIDE 91

Equations 1

Idea: Express equations over {∨, ·, 1} at the frame level. For an equation ε over {∨, ·, 1} we distribute products over joins to get s1 ∨ · · · ∨ sm = t1 ∨ · · · ∨ tn. si, tj: monoid terms. s1 ∨ · · · ∨ sm ≤ t1 ∨ · · · ∨ tn and t1 ∨ · · · ∨ tn ≤ s1 ∨ · · · ∨ sm. The first is equivalent to: &(sj ≤ t1 ∨ · · · ∨ tn). We proceed by example: x2y ≤ xy ∨ yx (x1 ∨ x2)2y ≤ (x1 ∨ x2)y ∨ y(x1 ∨ x2) x2

1y ∨ x1x2y ∨ x2x1y ∨ x2 2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2

x1x2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2 x1y ≤ v x2y ≤ v yx1 ≤ v yx2 ≤ v x1x2y ≤ v x1 ◦ y N z x2 ◦ y N z y ◦ x1 N z y ◦ x2 N z x1 ◦ x2 ◦ y N z R(ε)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 24 / 39

slide-92
SLIDE 92

Equations 1

Idea: Express equations over {∨, ·, 1} at the frame level. For an equation ε over {∨, ·, 1} we distribute products over joins to get s1 ∨ · · · ∨ sm = t1 ∨ · · · ∨ tn. si, tj: monoid terms. s1 ∨ · · · ∨ sm ≤ t1 ∨ · · · ∨ tn and t1 ∨ · · · ∨ tn ≤ s1 ∨ · · · ∨ sm. The first is equivalent to: &(sj ≤ t1 ∨ · · · ∨ tn). We proceed by example: x2y ≤ xy ∨ yx (x1 ∨ x2)2y ≤ (x1 ∨ x2)y ∨ y(x1 ∨ x2) x2

1y ∨ x1x2y ∨ x2x1y ∨ x2 2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2

x1x2y ≤ x1y ∨ x2y ∨ yx1 ∨ yx2 x1y ≤ v x2y ≤ v yx1 ≤ v yx2 ≤ v x1x2y ≤ v x1 ◦ y N z x2 ◦ y N z y ◦ x1 N z y ◦ x2 N z x1 ◦ x2 ◦ y N z R(ε)

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 24 / 39

slide-93
SLIDE 93

Equations 2

  • Theorem. If (W, B) is a Gentzen frame and ε an equation over {∨, ·, 1},

then (W, B) satisfies R(ε) iff W+ satisfies ε. (The linearity of the denominator of R(ε) plays an important role in the proof.)

  • Corollary. If an equation over {∨, ·, 1} is valid in A, then it is also valid in

W+

A,B, for every partial subalgebra B of A.

Consequently, W+

A,B ∈ V.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 25 / 39

slide-94
SLIDE 94

Equations 2

  • Theorem. If (W, B) is a Gentzen frame and ε an equation over {∨, ·, 1},

then (W, B) satisfies R(ε) iff W+ satisfies ε. (The linearity of the denominator of R(ε) plays an important role in the proof.)

  • Corollary. If an equation over {∨, ·, 1} is valid in A, then it is also valid in

W+

A,B, for every partial subalgebra B of A.

Consequently, W+

A,B ∈ V.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 25 / 39

slide-95
SLIDE 95

Equations 2

  • Theorem. If (W, B) is a Gentzen frame and ε an equation over {∨, ·, 1},

then (W, B) satisfies R(ε) iff W+ satisfies ε. (The linearity of the denominator of R(ε) plays an important role in the proof.)

  • Corollary. If an equation over {∨, ·, 1} is valid in A, then it is also valid in

W+

A,B, for every partial subalgebra B of A.

Consequently, W+

A,B ∈ V.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 25 / 39

slide-96
SLIDE 96

Structural rules

Given an equation ε of the form t0 ≤ t1 ∨ · · · ∨ tn, where ti are {·, 1}-terms we construct the rule R(ε) u[t1] ⇒ a · · · u[tn] ⇒ a u[t0] ⇒ a (R(ε)) where the ti’s are evaluated in (W , ◦, ε). Such a rule is called linear if all variables in t0 are distinct.

  • Theorem. Every system obtained from FL by adding linear rules has the

cut elimination property. A set of rules of the form R(ε) is called reducing if there is a complexity measure that decreases with upward applications of the rules (and the rules of FL).

  • Theorem. Every system obtained from FL by adding linear reducing rules

is decidable. The subvariety of residuated lattices axiomatized by the corresponding equations has decidable equational theory.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 26 / 39

slide-97
SLIDE 97

Structural rules

Given an equation ε of the form t0 ≤ t1 ∨ · · · ∨ tn, where ti are {·, 1}-terms we construct the rule R(ε) u[t1] ⇒ a · · · u[tn] ⇒ a u[t0] ⇒ a (R(ε)) where the ti’s are evaluated in (W , ◦, ε). Such a rule is called linear if all variables in t0 are distinct.

  • Theorem. Every system obtained from FL by adding linear rules has the

cut elimination property. A set of rules of the form R(ε) is called reducing if there is a complexity measure that decreases with upward applications of the rules (and the rules of FL).

  • Theorem. Every system obtained from FL by adding linear reducing rules

is decidable. The subvariety of residuated lattices axiomatized by the corresponding equations has decidable equational theory.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 26 / 39

slide-98
SLIDE 98

Structural rules

Given an equation ε of the form t0 ≤ t1 ∨ · · · ∨ tn, where ti are {·, 1}-terms we construct the rule R(ε) u[t1] ⇒ a · · · u[tn] ⇒ a u[t0] ⇒ a (R(ε)) where the ti’s are evaluated in (W , ◦, ε). Such a rule is called linear if all variables in t0 are distinct.

  • Theorem. Every system obtained from FL by adding linear rules has the

cut elimination property. A set of rules of the form R(ε) is called reducing if there is a complexity measure that decreases with upward applications of the rules (and the rules of FL).

  • Theorem. Every system obtained from FL by adding linear reducing rules

is decidable. The subvariety of residuated lattices axiomatized by the corresponding equations has decidable equational theory.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 26 / 39

slide-99
SLIDE 99

Structural rules

Given an equation ε of the form t0 ≤ t1 ∨ · · · ∨ tn, where ti are {·, 1}-terms we construct the rule R(ε) u[t1] ⇒ a · · · u[tn] ⇒ a u[t0] ⇒ a (R(ε)) where the ti’s are evaluated in (W , ◦, ε). Such a rule is called linear if all variables in t0 are distinct.

  • Theorem. Every system obtained from FL by adding linear rules has the

cut elimination property. A set of rules of the form R(ε) is called reducing if there is a complexity measure that decreases with upward applications of the rules (and the rules of FL).

  • Theorem. Every system obtained from FL by adding linear reducing rules

is decidable. The subvariety of residuated lattices axiomatized by the corresponding equations has decidable equational theory.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 26 / 39

slide-100
SLIDE 100

Applications

Cut-elimination (CE) and finite model property (FMP) for FL and (cyclic) InFL. Generation by finite members for RL, InFL

  • M. Kozak 2008 proved distributive FL has the FMP, and using our

approach the same result holds for any extension of DFL with linear reducing structural rules The finite embeddability property (FEP) for integral RL with {∨, ·, 1}-axioms The above extend to the non-associative case, also with the addition

  • f suitable structural rules

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 27 / 39

slide-101
SLIDE 101

Applications

Cut-elimination (CE) and finite model property (FMP) for FL and (cyclic) InFL. Generation by finite members for RL, InFL

  • M. Kozak 2008 proved distributive FL has the FMP, and using our

approach the same result holds for any extension of DFL with linear reducing structural rules The finite embeddability property (FEP) for integral RL with {∨, ·, 1}-axioms The above extend to the non-associative case, also with the addition

  • f suitable structural rules

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 27 / 39

slide-102
SLIDE 102

Applications

Cut-elimination (CE) and finite model property (FMP) for FL and (cyclic) InFL. Generation by finite members for RL, InFL

  • M. Kozak 2008 proved distributive FL has the FMP, and using our

approach the same result holds for any extension of DFL with linear reducing structural rules The finite embeddability property (FEP) for integral RL with {∨, ·, 1}-axioms The above extend to the non-associative case, also with the addition

  • f suitable structural rules

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 27 / 39

slide-103
SLIDE 103

Applications

Cut-elimination (CE) and finite model property (FMP) for FL and (cyclic) InFL. Generation by finite members for RL, InFL

  • M. Kozak 2008 proved distributive FL has the FMP, and using our

approach the same result holds for any extension of DFL with linear reducing structural rules The finite embeddability property (FEP) for integral RL with {∨, ·, 1}-axioms The above extend to the non-associative case, also with the addition

  • f suitable structural rules

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 27 / 39

slide-104
SLIDE 104

(Un)decidability

  • Theorem. The quasiequational theory of RL is undecidable. (Because we

can embed semigroups/monoids.) The same holds for commutative RL. A lattice is modular if x ∧ (y ∨ (x ∧ z)) = (x ∧ y) ∨ (x ∧ z)

  • Theorem. The equational theory of modular RL is undecidable. (Using

the corresponding result for modular lattices by Freese 1980).

  • Theorem. The equational theory of commutative, distributive RL is

decidable (Galatos Raftery 2004, from decidability of relevant logic RW by Brady 1990).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 28 / 39

slide-105
SLIDE 105

(Un)decidability

  • Theorem. The quasiequational theory of RL is undecidable. (Because we

can embed semigroups/monoids.) The same holds for commutative RL. A lattice is modular if x ∧ (y ∨ (x ∧ z)) = (x ∧ y) ∨ (x ∧ z)

  • Theorem. The equational theory of modular RL is undecidable. (Using

the corresponding result for modular lattices by Freese 1980).

  • Theorem. The equational theory of commutative, distributive RL is

decidable (Galatos Raftery 2004, from decidability of relevant logic RW by Brady 1990).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 28 / 39

slide-106
SLIDE 106

(Un)decidability

  • Theorem. The quasiequational theory of RL is undecidable. (Because we

can embed semigroups/monoids.) The same holds for commutative RL. A lattice is modular if x ∧ (y ∨ (x ∧ z)) = (x ∧ y) ∨ (x ∧ z)

  • Theorem. The equational theory of modular RL is undecidable. (Using

the corresponding result for modular lattices by Freese 1980).

  • Theorem. The equational theory of commutative, distributive RL is

decidable (Galatos Raftery 2004, from decidability of relevant logic RW by Brady 1990).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 28 / 39

slide-107
SLIDE 107

(Un)decidability

  • Theorem. The quasiequational theory of RL is undecidable. (Because we

can embed semigroups/monoids.) The same holds for commutative RL. A lattice is modular if x ∧ (y ∨ (x ∧ z)) = (x ∧ y) ∨ (x ∧ z)

  • Theorem. The equational theory of modular RL is undecidable. (Using

the corresponding result for modular lattices by Freese 1980).

  • Theorem. The equational theory of commutative, distributive RL is

decidable (Galatos Raftery 2004, from decidability of relevant logic RW by Brady 1990).

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 28 / 39

slide-108
SLIDE 108

Word problem

A finitely presented algebra A = (X|R) (in a class K) has a solvable word problem (WP) if there is an algorithm that, given any pair of words over X, decides if they are equal or not. A class of algebras has solvable WP if all finitely presented algebras in it do. For example, the varieties of semigroups, groups, ℓ-groups, modular lattices have unsolvable WP. Theorem [Galatos 2002]: The variety CDRL of commutative, distributive residuated lattices has unsolvable WP.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 29 / 39

slide-109
SLIDE 109

Word problem

A finitely presented algebra A = (X|R) (in a class K) has a solvable word problem (WP) if there is an algorithm that, given any pair of words over X, decides if they are equal or not. A class of algebras has solvable WP if all finitely presented algebras in it do. For example, the varieties of semigroups, groups, ℓ-groups, modular lattices have unsolvable WP. Theorem [Galatos 2002]: The variety CDRL of commutative, distributive residuated lattices has unsolvable WP.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 29 / 39

slide-110
SLIDE 110

Word problem

A finitely presented algebra A = (X|R) (in a class K) has a solvable word problem (WP) if there is an algorithm that, given any pair of words over X, decides if they are equal or not. A class of algebras has solvable WP if all finitely presented algebras in it do. For example, the varieties of semigroups, groups, ℓ-groups, modular lattices have unsolvable WP. Theorem [Galatos 2002]: The variety CDRL of commutative, distributive residuated lattices has unsolvable WP.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 29 / 39

slide-111
SLIDE 111

Word problem

A finitely presented algebra A = (X|R) (in a class K) has a solvable word problem (WP) if there is an algorithm that, given any pair of words over X, decides if they are equal or not. A class of algebras has solvable WP if all finitely presented algebras in it do. For example, the varieties of semigroups, groups, ℓ-groups, modular lattices have unsolvable WP. Theorem [Galatos 2002]: The variety CDRL of commutative, distributive residuated lattices has unsolvable WP.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 29 / 39

slide-112
SLIDE 112

Word problem for CDRL is unsolvable

Main idea: Embed semigroups, whose WP is unsolvable. Residuated lattices have a semigroup operation ·, but commutative semigroups have a decidable WP. Alternative approach: Come up with another term definable operation ⊙ in commutative distributive residuated lattices that is associative and embeds all semigroups. Technique: Coordinization in projective geometry and modular lattices, developed by J. von Neumann for continuous geometries, and applied by R. Freese to modular lattices, A. Urquhart to relevance logics, H. Andreak, S. Givant, I. Nemeti to symmetric relation algebras, and N. Galatos to CDRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 30 / 39

slide-113
SLIDE 113

Word problem for CDRL is unsolvable

Main idea: Embed semigroups, whose WP is unsolvable. Residuated lattices have a semigroup operation ·, but commutative semigroups have a decidable WP. Alternative approach: Come up with another term definable operation ⊙ in commutative distributive residuated lattices that is associative and embeds all semigroups. Technique: Coordinization in projective geometry and modular lattices, developed by J. von Neumann for continuous geometries, and applied by R. Freese to modular lattices, A. Urquhart to relevance logics, H. Andreak, S. Givant, I. Nemeti to symmetric relation algebras, and N. Galatos to CDRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 30 / 39

slide-114
SLIDE 114

Word problem for CDRL is unsolvable

Main idea: Embed semigroups, whose WP is unsolvable. Residuated lattices have a semigroup operation ·, but commutative semigroups have a decidable WP. Alternative approach: Come up with another term definable operation ⊙ in commutative distributive residuated lattices that is associative and embeds all semigroups. Technique: Coordinization in projective geometry and modular lattices, developed by J. von Neumann for continuous geometries, and applied by R. Freese to modular lattices, A. Urquhart to relevance logics, H. Andreak, S. Givant, I. Nemeti to symmetric relation algebras, and N. Galatos to CDRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 30 / 39

slide-115
SLIDE 115

Word problem for CDRL is unsolvable

Main idea: Embed semigroups, whose WP is unsolvable. Residuated lattices have a semigroup operation ·, but commutative semigroups have a decidable WP. Alternative approach: Come up with another term definable operation ⊙ in commutative distributive residuated lattices that is associative and embeds all semigroups. Technique: Coordinization in projective geometry and modular lattices, developed by J. von Neumann for continuous geometries, and applied by R. Freese to modular lattices, A. Urquhart to relevance logics, H. Andreak, S. Givant, I. Nemeti to symmetric relation algebras, and N. Galatos to CDRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 30 / 39

slide-116
SLIDE 116

Undecidability of quasiequational theory

A quasi-equation is a formula of the form (s1 = t1 & s2 = t2 & · · · & sn = tn) ⇒ s = t The decidability of the quasi-equational theory states that there is an algorithm that decides all quasi-equations of the above form. The equivalent logical notion is the decidability of the deducibility relation for formulas. Corollary The quasi-equational theory of CDRL is undecidable. Hence CDRL does not have the FEP, although we saw earlier that the equational theory is decidable.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 31 / 39

slide-117
SLIDE 117

Undecidability of quasiequational theory

A quasi-equation is a formula of the form (s1 = t1 & s2 = t2 & · · · & sn = tn) ⇒ s = t The decidability of the quasi-equational theory states that there is an algorithm that decides all quasi-equations of the above form. The equivalent logical notion is the decidability of the deducibility relation for formulas. Corollary The quasi-equational theory of CDRL is undecidable. Hence CDRL does not have the FEP, although we saw earlier that the equational theory is decidable.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 31 / 39

slide-118
SLIDE 118

Undecidability of quasiequational theory

A quasi-equation is a formula of the form (s1 = t1 & s2 = t2 & · · · & sn = tn) ⇒ s = t The decidability of the quasi-equational theory states that there is an algorithm that decides all quasi-equations of the above form. The equivalent logical notion is the decidability of the deducibility relation for formulas. Corollary The quasi-equational theory of CDRL is undecidable. Hence CDRL does not have the FEP, although we saw earlier that the equational theory is decidable.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 31 / 39

slide-119
SLIDE 119

Undecidability of quasiequational theory

A quasi-equation is a formula of the form (s1 = t1 & s2 = t2 & · · · & sn = tn) ⇒ s = t The decidability of the quasi-equational theory states that there is an algorithm that decides all quasi-equations of the above form. The equivalent logical notion is the decidability of the deducibility relation for formulas. Corollary The quasi-equational theory of CDRL is undecidable. Hence CDRL does not have the FEP, although we saw earlier that the equational theory is decidable.

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 31 / 39

slide-120
SLIDE 120

Further results on decidability

  • C. Holland, S. H. McCleary 1979: ℓ-groups have decidable equation theory

A.M.W. Glass, Y. Gurevich 1983: ℓ-groups have undecidable word problem

  • N. G. Hisamiev 1966: abelian ℓ-groups have decidable universal theory, by
  • V. Weispfenning 1986, in fact co-NP-complete, but by Y. Gurevich 1967,

the first-order theory is hereditarily undecidable MV-algebras have FEP because of a connections to linear programming IGMV and GMV have decidable equational theory because of a connections to ℓ-groups (N. Galatos, C. Tsinakis 2004), but no FMP

  • P. Jipsen and F. Montagna 2006, 2008: GBL and IGBL do not have FMP

but normal GBL-algebras have FEP

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 32 / 39

slide-121
SLIDE 121

Further results on decidability

  • C. Holland, S. H. McCleary 1979: ℓ-groups have decidable equation theory

A.M.W. Glass, Y. Gurevich 1983: ℓ-groups have undecidable word problem

  • N. G. Hisamiev 1966: abelian ℓ-groups have decidable universal theory, by
  • V. Weispfenning 1986, in fact co-NP-complete, but by Y. Gurevich 1967,

the first-order theory is hereditarily undecidable MV-algebras have FEP because of a connections to linear programming IGMV and GMV have decidable equational theory because of a connections to ℓ-groups (N. Galatos, C. Tsinakis 2004), but no FMP

  • P. Jipsen and F. Montagna 2006, 2008: GBL and IGBL do not have FMP

but normal GBL-algebras have FEP

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 32 / 39

slide-122
SLIDE 122

Further results on decidability

  • C. Holland, S. H. McCleary 1979: ℓ-groups have decidable equation theory

A.M.W. Glass, Y. Gurevich 1983: ℓ-groups have undecidable word problem

  • N. G. Hisamiev 1966: abelian ℓ-groups have decidable universal theory, by
  • V. Weispfenning 1986, in fact co-NP-complete, but by Y. Gurevich 1967,

the first-order theory is hereditarily undecidable MV-algebras have FEP because of a connections to linear programming IGMV and GMV have decidable equational theory because of a connections to ℓ-groups (N. Galatos, C. Tsinakis 2004), but no FMP

  • P. Jipsen and F. Montagna 2006, 2008: GBL and IGBL do not have FMP

but normal GBL-algebras have FEP

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 32 / 39

slide-123
SLIDE 123

Further results on decidability

  • C. Holland, S. H. McCleary 1979: ℓ-groups have decidable equation theory

A.M.W. Glass, Y. Gurevich 1983: ℓ-groups have undecidable word problem

  • N. G. Hisamiev 1966: abelian ℓ-groups have decidable universal theory, by
  • V. Weispfenning 1986, in fact co-NP-complete, but by Y. Gurevich 1967,

the first-order theory is hereditarily undecidable MV-algebras have FEP because of a connections to linear programming IGMV and GMV have decidable equational theory because of a connections to ℓ-groups (N. Galatos, C. Tsinakis 2004), but no FMP

  • P. Jipsen and F. Montagna 2006, 2008: GBL and IGBL do not have FMP

but normal GBL-algebras have FEP

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 32 / 39

slide-124
SLIDE 124

Further results on decidability

  • C. Holland, S. H. McCleary 1979: ℓ-groups have decidable equation theory

A.M.W. Glass, Y. Gurevich 1983: ℓ-groups have undecidable word problem

  • N. G. Hisamiev 1966: abelian ℓ-groups have decidable universal theory, by
  • V. Weispfenning 1986, in fact co-NP-complete, but by Y. Gurevich 1967,

the first-order theory is hereditarily undecidable MV-algebras have FEP because of a connections to linear programming IGMV and GMV have decidable equational theory because of a connections to ℓ-groups (N. Galatos, C. Tsinakis 2004), but no FMP

  • P. Jipsen and F. Montagna 2006, 2008: GBL and IGBL do not have FMP

but normal GBL-algebras have FEP

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 32 / 39

slide-125
SLIDE 125

Further results on decidability

  • C. Holland, S. H. McCleary 1979: ℓ-groups have decidable equation theory

A.M.W. Glass, Y. Gurevich 1983: ℓ-groups have undecidable word problem

  • N. G. Hisamiev 1966: abelian ℓ-groups have decidable universal theory, by
  • V. Weispfenning 1986, in fact co-NP-complete, but by Y. Gurevich 1967,

the first-order theory is hereditarily undecidable MV-algebras have FEP because of a connections to linear programming IGMV and GMV have decidable equational theory because of a connections to ℓ-groups (N. Galatos, C. Tsinakis 2004), but no FMP

  • P. Jipsen and F. Montagna 2006, 2008: GBL and IGBL do not have FMP

but normal GBL-algebras have FEP

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 32 / 39

slide-126
SLIDE 126

Poset products

The poset product uses a partial order on the index set to define a subset

  • f the direct product.

Specifically, let X = (X, ≤) be a poset, and assume {Ai | i ∈ X} is a family of algebras that have two constant operations denoted 0, 1. The poset product of {Ai | i ∈ X} is

  • X

Ai = {f ∈

  • i∈X

Ai | f (i) = 0 or f (j) = 1 for all i < j in X} If X is an antichain then the poset product is the same as the direct product If X is a chain and the Ai are ordered, then the poset product is the (amalgamated) ordinal sum of the Ai

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 33 / 39

slide-127
SLIDE 127

Poset products

The poset product uses a partial order on the index set to define a subset

  • f the direct product.

Specifically, let X = (X, ≤) be a poset, and assume {Ai | i ∈ X} is a family of algebras that have two constant operations denoted 0, 1. The poset product of {Ai | i ∈ X} is

  • X

Ai = {f ∈

  • i∈X

Ai | f (i) = 0 or f (j) = 1 for all i < j in X} If X is an antichain then the poset product is the same as the direct product If X is a chain and the Ai are ordered, then the poset product is the (amalgamated) ordinal sum of the Ai

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 33 / 39

slide-128
SLIDE 128

Poset products

The poset product uses a partial order on the index set to define a subset

  • f the direct product.

Specifically, let X = (X, ≤) be a poset, and assume {Ai | i ∈ X} is a family of algebras that have two constant operations denoted 0, 1. The poset product of {Ai | i ∈ X} is

  • X

Ai = {f ∈

  • i∈X

Ai | f (i) = 0 or f (j) = 1 for all i < j in X} If X is an antichain then the poset product is the same as the direct product If X is a chain and the Ai are ordered, then the poset product is the (amalgamated) ordinal sum of the Ai

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 33 / 39

slide-129
SLIDE 129

Poset products

The poset product uses a partial order on the index set to define a subset

  • f the direct product.

Specifically, let X = (X, ≤) be a poset, and assume {Ai | i ∈ X} is a family of algebras that have two constant operations denoted 0, 1. The poset product of {Ai | i ∈ X} is

  • X

Ai = {f ∈

  • i∈X

Ai | f (i) = 0 or f (j) = 1 for all i < j in X} If X is an antichain then the poset product is the same as the direct product If X is a chain and the Ai are ordered, then the poset product is the (amalgamated) ordinal sum of the Ai

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 33 / 39

slide-130
SLIDE 130

Poset products

The poset product uses a partial order on the index set to define a subset

  • f the direct product.

Specifically, let X = (X, ≤) be a poset, and assume {Ai | i ∈ X} is a family of algebras that have two constant operations denoted 0, 1. The poset product of {Ai | i ∈ X} is

  • X

Ai = {f ∈

  • i∈X

Ai | f (i) = 0 or f (j) = 1 for all i < j in X} If X is an antichain then the poset product is the same as the direct product If X is a chain and the Ai are ordered, then the poset product is the (amalgamated) ordinal sum of the Ai

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 33 / 39

slide-131
SLIDE 131

For an ℓ-groupoid A define IA = {c ∈ A | c ∧ x = cx = xc for all x ∈ A} Note that ∧ distributes over ∨ in IA, but IA need not be a subalgebra of A A GBL-algebra is a residuated lattice that satisfies x ≤ y ⇒ x = (x/y)y = y(y\x) [J., Montagna 2006] prove that for bounded GBL-algebras, IA is a subalgebra, hence a Heyting algebra contained in A, and B(A) is the subalgebra of complemented elements of IA. For MV-algebras IA = B(A)

Lemma

Let A be a FLw-algebra and let a, b ∈ IA with a ≤ b. Then the interval [a, b] = {x ∈ A | a ≤ x ≤ b} is a FLw-algebra, with 0 = a, 1 = b, ∧, ∨, · inherited from A, and x\y = (x\Ay) ∧ b, x/y = (x/Ay) ∧ b. If A is a GBL-algebra, then so is [a, b].

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 34 / 39

slide-132
SLIDE 132

For an ℓ-groupoid A define IA = {c ∈ A | c ∧ x = cx = xc for all x ∈ A} Note that ∧ distributes over ∨ in IA, but IA need not be a subalgebra of A A GBL-algebra is a residuated lattice that satisfies x ≤ y ⇒ x = (x/y)y = y(y\x) [J., Montagna 2006] prove that for bounded GBL-algebras, IA is a subalgebra, hence a Heyting algebra contained in A, and B(A) is the subalgebra of complemented elements of IA. For MV-algebras IA = B(A)

Lemma

Let A be a FLw-algebra and let a, b ∈ IA with a ≤ b. Then the interval [a, b] = {x ∈ A | a ≤ x ≤ b} is a FLw-algebra, with 0 = a, 1 = b, ∧, ∨, · inherited from A, and x\y = (x\Ay) ∧ b, x/y = (x/Ay) ∧ b. If A is a GBL-algebra, then so is [a, b].

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 34 / 39

slide-133
SLIDE 133

For an ℓ-groupoid A define IA = {c ∈ A | c ∧ x = cx = xc for all x ∈ A} Note that ∧ distributes over ∨ in IA, but IA need not be a subalgebra of A A GBL-algebra is a residuated lattice that satisfies x ≤ y ⇒ x = (x/y)y = y(y\x) [J., Montagna 2006] prove that for bounded GBL-algebras, IA is a subalgebra, hence a Heyting algebra contained in A, and B(A) is the subalgebra of complemented elements of IA. For MV-algebras IA = B(A)

Lemma

Let A be a FLw-algebra and let a, b ∈ IA with a ≤ b. Then the interval [a, b] = {x ∈ A | a ≤ x ≤ b} is a FLw-algebra, with 0 = a, 1 = b, ∧, ∨, · inherited from A, and x\y = (x\Ay) ∧ b, x/y = (x/Ay) ∧ b. If A is a GBL-algebra, then so is [a, b].

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 34 / 39

slide-134
SLIDE 134

For an ℓ-groupoid A define IA = {c ∈ A | c ∧ x = cx = xc for all x ∈ A} Note that ∧ distributes over ∨ in IA, but IA need not be a subalgebra of A A GBL-algebra is a residuated lattice that satisfies x ≤ y ⇒ x = (x/y)y = y(y\x) [J., Montagna 2006] prove that for bounded GBL-algebras, IA is a subalgebra, hence a Heyting algebra contained in A, and B(A) is the subalgebra of complemented elements of IA. For MV-algebras IA = B(A)

Lemma

Let A be a FLw-algebra and let a, b ∈ IA with a ≤ b. Then the interval [a, b] = {x ∈ A | a ≤ x ≤ b} is a FLw-algebra, with 0 = a, 1 = b, ∧, ∨, · inherited from A, and x\y = (x\Ay) ∧ b, x/y = (x/Ay) ∧ b. If A is a GBL-algebra, then so is [a, b].

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 34 / 39

slide-135
SLIDE 135

For an ℓ-groupoid A define IA = {c ∈ A | c ∧ x = cx = xc for all x ∈ A} Note that ∧ distributes over ∨ in IA, but IA need not be a subalgebra of A A GBL-algebra is a residuated lattice that satisfies x ≤ y ⇒ x = (x/y)y = y(y\x) [J., Montagna 2006] prove that for bounded GBL-algebras, IA is a subalgebra, hence a Heyting algebra contained in A, and B(A) is the subalgebra of complemented elements of IA. For MV-algebras IA = B(A)

Lemma

Let A be a FLw-algebra and let a, b ∈ IA with a ≤ b. Then the interval [a, b] = {x ∈ A | a ≤ x ≤ b} is a FLw-algebra, with 0 = a, 1 = b, ∧, ∨, · inherited from A, and x\y = (x\Ay) ∧ b, x/y = (x/Ay) ∧ b. If A is a GBL-algebra, then so is [a, b].

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 34 / 39

slide-136
SLIDE 136

For an ℓ-groupoid A define IA = {c ∈ A | c ∧ x = cx = xc for all x ∈ A} Note that ∧ distributes over ∨ in IA, but IA need not be a subalgebra of A A GBL-algebra is a residuated lattice that satisfies x ≤ y ⇒ x = (x/y)y = y(y\x) [J., Montagna 2006] prove that for bounded GBL-algebras, IA is a subalgebra, hence a Heyting algebra contained in A, and B(A) is the subalgebra of complemented elements of IA. For MV-algebras IA = B(A)

Lemma

Let A be a FLw-algebra and let a, b ∈ IA with a ≤ b. Then the interval [a, b] = {x ∈ A | a ≤ x ≤ b} is a FLw-algebra, with 0 = a, 1 = b, ∧, ∨, · inherited from A, and x\y = (x\Ay) ∧ b, x/y = (x/Ay) ∧ b. If A is a GBL-algebra, then so is [a, b].

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 34 / 39

slide-137
SLIDE 137

We now generalize the poset sum decomposition result of [J., Montagna 2006] from finite GBL-algebras to certain FLw-algebras

Theorem

Consider a FLw-algebra A with a finite subalgebra C such that C ⊆ IA, and let X be the dual of the partially ordered set of completely join irreducible elements of C. For c ∈ X, let c∗ denote the unique lower cover of c in C. If Ac = ↓c∗ ⊕ [c∗, c] for all c ∈ X then A ∼ =

  • X

[c∗, c]. The condition Ac = ↓c∗ ⊕ [c∗, c] is actually satisfied for many GBL-algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 35 / 39

slide-138
SLIDE 138

We now generalize the poset sum decomposition result of [J., Montagna 2006] from finite GBL-algebras to certain FLw-algebras

Theorem

Consider a FLw-algebra A with a finite subalgebra C such that C ⊆ IA, and let X be the dual of the partially ordered set of completely join irreducible elements of C. For c ∈ X, let c∗ denote the unique lower cover of c in C. If Ac = ↓c∗ ⊕ [c∗, c] for all c ∈ X then A ∼ =

  • X

[c∗, c]. The condition Ac = ↓c∗ ⊕ [c∗, c] is actually satisfied for many GBL-algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 35 / 39

slide-139
SLIDE 139

We now generalize the poset sum decomposition result of [J., Montagna 2006] from finite GBL-algebras to certain FLw-algebras

Theorem

Consider a FLw-algebra A with a finite subalgebra C such that C ⊆ IA, and let X be the dual of the partially ordered set of completely join irreducible elements of C. For c ∈ X, let c∗ denote the unique lower cover of c in C. If Ac = ↓c∗ ⊕ [c∗, c] for all c ∈ X then A ∼ =

  • X

[c∗, c]. The condition Ac = ↓c∗ ⊕ [c∗, c] is actually satisfied for many GBL-algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 35 / 39

slide-140
SLIDE 140

A GBL-algebra is normal if every filter is a normal filter

Theorem (J., Montagna)

A Blok-Ferreirim decomposition for GBL-algebras: Every subdirectly irreducible normal integral GBL-algebra decomposes as the ordinal sum of an integral GBL-algebra and a linearly ordered integral GMV-algebra. A residuated lattice is n-potent if it satisfies xn+1 = xn [J., Montagna] prove that any n-potent GBL-algebra is commutative, hence normal, so e.g. any finite GBL-algebra is commutative

Corollary

Suppose A is an integral normal GBL-algebra such that IA is finite Then A is isomorphic to a poset product of linearly ordered IGMV-algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 36 / 39

slide-141
SLIDE 141

A GBL-algebra is normal if every filter is a normal filter

Theorem (J., Montagna)

A Blok-Ferreirim decomposition for GBL-algebras: Every subdirectly irreducible normal integral GBL-algebra decomposes as the ordinal sum of an integral GBL-algebra and a linearly ordered integral GMV-algebra. A residuated lattice is n-potent if it satisfies xn+1 = xn [J., Montagna] prove that any n-potent GBL-algebra is commutative, hence normal, so e.g. any finite GBL-algebra is commutative

Corollary

Suppose A is an integral normal GBL-algebra such that IA is finite Then A is isomorphic to a poset product of linearly ordered IGMV-algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 36 / 39

slide-142
SLIDE 142

A GBL-algebra is normal if every filter is a normal filter

Theorem (J., Montagna)

A Blok-Ferreirim decomposition for GBL-algebras: Every subdirectly irreducible normal integral GBL-algebra decomposes as the ordinal sum of an integral GBL-algebra and a linearly ordered integral GMV-algebra. A residuated lattice is n-potent if it satisfies xn+1 = xn [J., Montagna] prove that any n-potent GBL-algebra is commutative, hence normal, so e.g. any finite GBL-algebra is commutative

Corollary

Suppose A is an integral normal GBL-algebra such that IA is finite Then A is isomorphic to a poset product of linearly ordered IGMV-algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 36 / 39

slide-143
SLIDE 143

A GBL-algebra is normal if every filter is a normal filter

Theorem (J., Montagna)

A Blok-Ferreirim decomposition for GBL-algebras: Every subdirectly irreducible normal integral GBL-algebra decomposes as the ordinal sum of an integral GBL-algebra and a linearly ordered integral GMV-algebra. A residuated lattice is n-potent if it satisfies xn+1 = xn [J., Montagna] prove that any n-potent GBL-algebra is commutative, hence normal, so e.g. any finite GBL-algebra is commutative

Corollary

Suppose A is an integral normal GBL-algebra such that IA is finite Then A is isomorphic to a poset product of linearly ordered IGMV-algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 36 / 39

slide-144
SLIDE 144

A GBL-algebra is normal if every filter is a normal filter

Theorem (J., Montagna)

A Blok-Ferreirim decomposition for GBL-algebras: Every subdirectly irreducible normal integral GBL-algebra decomposes as the ordinal sum of an integral GBL-algebra and a linearly ordered integral GMV-algebra. A residuated lattice is n-potent if it satisfies xn+1 = xn [J., Montagna] prove that any n-potent GBL-algebra is commutative, hence normal, so e.g. any finite GBL-algebra is commutative

Corollary

Suppose A is an integral normal GBL-algebra such that IA is finite Then A is isomorphic to a poset product of linearly ordered IGMV-algebras

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 36 / 39

slide-145
SLIDE 145

Open Problems

Do cancellative residuated lattices have a decidable equational theory or a cut free Gentzen system? Do (I)GBL-algebras have a decidable equational theory or a cut free Gentzen system? Is provability in FLc decidable, i.e. does the variety FLc have a decidable equational theory? A cut-free Gentzen system is known Develop a structure theory for infinite IGBL-algebras Do commutative cancellative residuated lattices satisfy any lattice equations that do not hold in all lattices? Investigate the structure of free residuated lattices. Even the 1-generated case is not well understood Find practical decision procedures for deducibility in FL(e)w or for deciding quasiequations in (C)IRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 37 / 39

slide-146
SLIDE 146

Open Problems

Do cancellative residuated lattices have a decidable equational theory or a cut free Gentzen system? Do (I)GBL-algebras have a decidable equational theory or a cut free Gentzen system? Is provability in FLc decidable, i.e. does the variety FLc have a decidable equational theory? A cut-free Gentzen system is known Develop a structure theory for infinite IGBL-algebras Do commutative cancellative residuated lattices satisfy any lattice equations that do not hold in all lattices? Investigate the structure of free residuated lattices. Even the 1-generated case is not well understood Find practical decision procedures for deducibility in FL(e)w or for deciding quasiequations in (C)IRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 37 / 39

slide-147
SLIDE 147

Open Problems

Do cancellative residuated lattices have a decidable equational theory or a cut free Gentzen system? Do (I)GBL-algebras have a decidable equational theory or a cut free Gentzen system? Is provability in FLc decidable, i.e. does the variety FLc have a decidable equational theory? A cut-free Gentzen system is known Develop a structure theory for infinite IGBL-algebras Do commutative cancellative residuated lattices satisfy any lattice equations that do not hold in all lattices? Investigate the structure of free residuated lattices. Even the 1-generated case is not well understood Find practical decision procedures for deducibility in FL(e)w or for deciding quasiequations in (C)IRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 37 / 39

slide-148
SLIDE 148

Open Problems

Do cancellative residuated lattices have a decidable equational theory or a cut free Gentzen system? Do (I)GBL-algebras have a decidable equational theory or a cut free Gentzen system? Is provability in FLc decidable, i.e. does the variety FLc have a decidable equational theory? A cut-free Gentzen system is known Develop a structure theory for infinite IGBL-algebras Do commutative cancellative residuated lattices satisfy any lattice equations that do not hold in all lattices? Investigate the structure of free residuated lattices. Even the 1-generated case is not well understood Find practical decision procedures for deducibility in FL(e)w or for deciding quasiequations in (C)IRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 37 / 39

slide-149
SLIDE 149

Open Problems

Do cancellative residuated lattices have a decidable equational theory or a cut free Gentzen system? Do (I)GBL-algebras have a decidable equational theory or a cut free Gentzen system? Is provability in FLc decidable, i.e. does the variety FLc have a decidable equational theory? A cut-free Gentzen system is known Develop a structure theory for infinite IGBL-algebras Do commutative cancellative residuated lattices satisfy any lattice equations that do not hold in all lattices? Investigate the structure of free residuated lattices. Even the 1-generated case is not well understood Find practical decision procedures for deducibility in FL(e)w or for deciding quasiequations in (C)IRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 37 / 39

slide-150
SLIDE 150

Open Problems

Do cancellative residuated lattices have a decidable equational theory or a cut free Gentzen system? Do (I)GBL-algebras have a decidable equational theory or a cut free Gentzen system? Is provability in FLc decidable, i.e. does the variety FLc have a decidable equational theory? A cut-free Gentzen system is known Develop a structure theory for infinite IGBL-algebras Do commutative cancellative residuated lattices satisfy any lattice equations that do not hold in all lattices? Investigate the structure of free residuated lattices. Even the 1-generated case is not well understood Find practical decision procedures for deducibility in FL(e)w or for deciding quasiequations in (C)IRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 37 / 39

slide-151
SLIDE 151

Open Problems

Do cancellative residuated lattices have a decidable equational theory or a cut free Gentzen system? Do (I)GBL-algebras have a decidable equational theory or a cut free Gentzen system? Is provability in FLc decidable, i.e. does the variety FLc have a decidable equational theory? A cut-free Gentzen system is known Develop a structure theory for infinite IGBL-algebras Do commutative cancellative residuated lattices satisfy any lattice equations that do not hold in all lattices? Investigate the structure of free residuated lattices. Even the 1-generated case is not well understood Find practical decision procedures for deducibility in FL(e)w or for deciding quasiequations in (C)IRL

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 37 / 39

slide-152
SLIDE 152

References I

  • F. Belardinelli, H. Ono and P. Jipsen, Algebraic aspects of cut elimination, Studia Logics

68 (2001), 1-32

  • W. Blok and C. van Alten, On the finite embeddability property for residuated lattices,

pocrims and BCK-algebras, Algebra Universalis, No. 48 (2002), 253–271

  • W. Blok and C. van Alten, On the finite embeddability property for residuated ordered

groupoids, Transactions of the AMS, 356(10) (2005), 4141–4157

  • A. Ciabattoni, N. Galatos and K. Terui. From Axioms to analytic rules in nonclassical

logics, LICS’08, 229–240

  • A. Ciabattoni, N. Galatos and K. Terui. The expressive power of structural rules for FL,

manuscript

  • N. Galatos and P. Jipsen. Residuated frames and applications to decidability, preprint
  • N. Galatos, P. Jipsen, T. Kowalski and H. Ono. Residuated Lattices: an algebraic

glimpse at substructural logics, Studies in Logics and the Foundations of Mathematics, Elsevier, 2007

  • N. Galatos and H. Ono. Algebraization, parameterized local deduction theorem and

interpolation for substructural logics over FL, Studia Logica 83 (2006), 279–308

  • N. Galatos and H. Ono. Cut elimination and strong separation for non-associative

substructural logics, manuscript

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 38 / 39

slide-153
SLIDE 153

References II

  • N. Galatos and J. Raftery, Adding involution to residuated structures, Studia Logica 77

(2004), 181–207

  • P. Jipsen and C. Tsinakis. A survey of residuated lattices, in ”Ordered Algebraic

Structures” (J. Martinez, editor), Kluwer Academic Publishers, Dordrecht, 2002, 19–56

  • P. Jipsen and F. Montagna, On the structure of generalized BL-algebras, Algebra

Universalis, 55 (2006), 226–237

  • P. Jipsen and F. Montagna, The Blok-Ferreirim theorem for normal GBL-algebras and

its application, Algebra Universalis, 60, to appear

  • P. Jipsen, Generalizations of Boolean products for lattice-ordered algebras, Annals of

Pure and Applied Logic, to appear

  • K. Terui, Which Structural Rules Admit Cut Elimination? - An Algebraic Criterion, JSL

72(3) (2007), 738–754

P Jipsen (Chapman), N Galatos (DU) Semantics: Residuated Frames ICLA, January, 2009 39 / 39