Linear Logic
Reference: Girard, J.-Y. (1987). Linear Logic. Theoretical Computer Science, 50(1), pp. 1–101. DOI · ScienceDirect (open archive)
Summary
Girard introduces linear logic by stripping the structural rules of weakening and contraction from classical and intuitionistic sequent calculus, then re-introducing them in a controlled way through exponential modalities ! (“of-course”) and ? (“why-not”). The resulting system is resource-aware: a hypothesis can be used exactly once unless explicitly marked reusable. The connectives split into two families — multiplicatives ⊗ (tensor, “with both available”) and ⅋ (par, “the disjunction with both still possible”) and additives & (with, “I choose”) and ⊕ (plus, “the world chooses”) — together with their units, negation as involution, and exponentials. The technical heart is proof nets: a graph-theoretic syntax that quotients away the inessential commutation steps of sequent calculus and gives a strongly normalising calculus of proofs. Cut elimination produces an algorithmic dynamic on proof nets that is the linear analogue of β-reduction. The deep consequences run in two directions: (i) classical/intuitionistic logic embeds into linear logic via the exponentials, so linear logic is a refinement, not a competitor; (ii) the multiplicatives correspond directly to concurrent communication (a two-place tensor is a synchronisation point), making linear logic the formal substrate of a Curry–Howard correspondence for Session Types and π-calculus communication. The paper inaugurated 35+ years of substructural logic, type-theoretic concurrency, and resource-typed programming.
Key Ideas
- Resources, not propositions: a linear hypothesis is used, not just consumed in a proof — closer to physical resources or messages than to mathematical truth.
- Multiplicatives vs additives:
A ⊗ Bmeans “I have both A and B” (independent resources);A & Bmeans “I have a choice between A and B but only one”;A ⅋ Bis the multiplicative dual to⊗;A ⊕ Bis the additive sum. - Exponentials reintroduce structural rules locally:
!Ais a reusable A; weakening and contraction apply to formulas under!;?Ais its de Morgan dual. - Involutive negation:
(A^⊥)^⊥ = Asyntactically; classical-style symmetry without classical-style proof multiplicities. - Proof nets: cut-elimination becomes graph rewriting; “essentially the same” sequent-calculus proofs become literally the same net.
- Curry–Howard for concurrency: linear-logic propositions ⇔ session types; proofs ⇔ processes; cut elimination ⇔ communication. The Caires–Pfenning correspondence (2010) and Wadler’s Propositions as Sessions (2012) turn this into a typed Pi-Calculus in which deadlock-freedom and session fidelity follow from cut elimination.
- Light fragments and complexity: Girard’s later light/elementary linear logics (LLL, ELL) capture polynomial-time and elementary computation respectively — proof-theoretic characterisations of complexity classes.
Connections
- Pi-Calculus
- A Calculus of Mobile Processes
- Session Types
- Multiparty Asynchronous Session Types
- Curry-Howard Correspondence
- Substructural Logic
- Proof Nets
- Multiplicative-Additive
- Exponential Modalities
- Process Calculi
- Algebraic Effects
- Robust Composition - Towards a Unified Approach to Access Control and Concurrency Control
Conceptual Contribution
- Claim: Removing weakening and contraction from sequent calculus and reintroducing them through controlled exponential modalities yields a resource-aware logic that is simultaneously a refinement of classical/intuitionistic logic and the natural type discipline for concurrent, resource-sensitive computation.
- Mechanism: Sequent calculus over multiplicative (
⊗,⅋), additive (&,⊕), and exponential (!,?) connectives with involutive negation and per-rule resource discipline; proof nets as cut-elimination-amenable graphical syntax; cut elimination as the dynamics of proofs; embeddings of classical and intuitionistic logic via exponentials. - Concepts introduced/used: Linear Logic, Multiplicative-Additive, Exponential Modalities, Proof Nets, Substructural Logic, Curry-Howard Correspondence (concurrent extension).
- Stance: foundational technical paper.
- Relates to: Provides the type-theoretic substrate that turns the Pi-Calculus from an untyped calculus into a typed one whose well-typed processes are deadlock-free and session-faithful — Caires & Pfenning (2010) and Wadler (2012) make this explicit. Every later session-typed language (Multiparty Asynchronous Session Types, the choreographic-programming languages of Structured Communication-Centred Programming for Web Services and Pact - A Choreographic Language for Agentic Ecosystems) inherits this connection. Outside concurrency, linear logic underlies Rust-style ownership types, region-based memory management, separation logic, and the linear-types extension that landed in Haskell — all engineering descendants of Girard’s resource discipline. Conceptually relevant to Capability Security: a capability is a linear resource (forgeable copies break it), and substructural type systems that prevent capability duplication realise the Dennis–Van Horn ideal in a typed setting; this connection is exploited in modern capability-typed languages (Pony’s reference capabilities, Project Verona, Rust’s borrow checker).
Tags
#linear-logic #girard #substructural #proof-theory #foundations #curry-howard #session-types-foundation