Focused Proofs in Sequent Calculi

• Focused proofs are a refined framework that organizes inferences into negative (invertible) and positive (focused) phases to reduce nondeterminism.
• They employ explicit polarity assignments and phase shifts, enabling systematic application of invertible rules and controlled focus on non-invertible steps.
• This methodology underpins efficient automation in proof assistants and facilitates proof normalization, cut elimination, and goal-directed search strategies.

Focused proofs are a refined class of proof objects in sequent calculi, characterized by the explicit structuring of inferences into phases based on the invertibility of rules and the polarity of formulas. The central goal of focused proof systems is to minimize proof search nondeterminism by organizing derivations so that invertible (asynchronous/negative) rules are applied eagerly in bulk, whereas non-invertible (synchronous/positive) rules are grouped into explicit focus phases. This discipline underpins proof normalization, goal-directed search procedures, and efficient automation in proof assistants. The technique originated in linear logic but has been extensively developed for intuitionistic, classical, and substructural logics, including the logic of Bunched Implications (BI) (Gheorghiu et al., 2020, Simmons, 2011, 0708.2252, Nigam et al., 2015, Schwartzentruber et al., 2023).

1. Polarity, Invertibility, and the Focusing Discipline

In a focused calculus, each atomic formula is given a polarity (positive or negative), and compound formulas inherit polarity accordingly. For example, positive formulas typically include positive atoms, disjunctions, and other connectives whose introduction rules are non-invertible, while negative formulas include negative atoms, conjunctions, and implications with invertible right-rules. Polarity-shifting operators (“upshift” and “downshift”) are included to mediate between phases but carry no logical content (Gheorghiu et al., 2020).

Invertible rules (negative phase) are those where, for any derivable sequent, the premises induced by the rule are also derivable, eliminating backtracking. Non-invertible rules (positive phase) require committing to a choice, such as selecting a disjunct or existential witness, and must occur under focus (0708.2252, Simmons, 2011).

Focused derivations enforce a two-phase alternation: all invertible rules are applied maximally in the negative phase; then, a single principal formula is chosen and decomposed by its non-invertible rules in the positive (focus) phase (Simmons, 2011, Nigam et al., 2015).

2. Focused Sequent Calculi: Syntax and Rules

The focused sequent calculus typically uses sequents and rule-sets that explicitly encode phase data. In intuitionistic logic (LJF, LJF-polarity), sequents take the form $\Gamma \longrightarrow R$ (asynchronous/unfocused) or $\Gamma \Rightarrow P$ (synchronous/focused), where $\Gamma$ is a multiset of negative formulas and positive atoms, and $P$ is positive (0708.2252).

In BI and substructural logics, contexts may be “bunches” or “nested multisets” that represent combinations of additive and multiplicative structure. The focused calculus employs dual sorts of nests and a pairing with the formula under proof (Gheorghiu et al., 2020). Focusing phases are made explicit with distinguished rules for shifting between negative and positive phases and decomposing focused formulas both on the left and on the right.

The following table summarizes the association of connective types and rule invertibility (example from BI):

Connective	Phase (Rule Type)	Typical Example
$\wedge^-$, $\to$, $\parr$, $\top^-$	Negative (invertible)	Apply right-rule in negative phase
$\vee$, $*$, $\Gamma \Rightarrow P$0, $\Gamma \Rightarrow P$1	Positive (non-invertible)	Decompose via focus

Invertible rules are permitted freely in the negative phase, while positive rules occur synchronously within a focus (Gheorghiu et al., 2020, Nigam et al., 2015).

3. Metatheorems: Cut Elimination and Focalization

Cut admissibility and focalization are the primary metatheorems for focused calculi. Cut admissibility states that any cut applied to focused sequents can be eliminated—often by structural induction without the need for auxiliary invertibility lemmas—yielding cut-free focused proofs (Simmons, 2011, Gheorghiu et al., 2020).

The focalization theorem establishes that every cut-free, non-focused proof (in a standard sequent calculus) can be transformed into a cut-free focused proof under any polarization strategy. The proof relies on two “internal” metatheorems: cut admissibility (internal soundness) and identity expansion (internal completeness—showing every identity axiom has a focused derivation for all positive and negative formulas) (Simmons, 2011). This mechanism enables not only normalization (ensuring unique normal-form derivations per sequent, up to minute choices of focus) but also significantly reduces the nondeterminism in backward proof search.

4. Automation, Permutation Graphs, and Synthesis of Focused Systems

The construction of focused proof systems can be automated by analyzing the permutation structure of rules. The permutation graph of a sequent system encodes, for each pair of inference rules $\Gamma \Rightarrow P$2, whether they permute: $\Gamma \Rightarrow P$3 if, in every proof, applications of $\Gamma \Rightarrow P$4 above $\Gamma \Rightarrow P$5 can be swapped. Partitioning this graph into “negative” (all mutually permutable, invertible) and “positive” (non-invertible) cliques yields a canonical focused system (Nigam et al., 2015).

This approach supports the automated synthesis of focused versions of LJ (LJF), LK (classical logic), or linear systems (MALLF) and extends, with refinement, to BI and other substructural frameworks. Admissibility of contraction is inferred from the permutation graph: contraction is admissible in a phase if it permutes up all the rules of that phase. Limitations appear in calculi with highly non-affine or ordered contexts (Nigam et al., 2015).

5. Impact: Proof Search, Normalization, and Meta-theoretic Automation

Focused proofs drastically reduce proof search nondeterminism. In proof assistants and logic programming, the two-phase structure enables deterministic application of invertible rules, with search space explosion controlled by confining nondeterminism to focus selection (Gheorghiu et al., 2020, 0708.2252). This principle underlies effective goal-directed proof strategies and justifies implementations in proof assistants such as Beluga, where focused proof search is implemented as tactics (e.g., “auto” in Harpoon), enabling the automatic or semi-automatic discharge of routine meta-theoretic lemmas (Schwartzentruber et al., 2023).

Normalization—i.e., the representation of every cut-free proof as a focused (phase-alternating) derivation—also reveals connections to operational semantics, especially in type theories, compiling proof search patterns, and logic programming (Simmons, 2011, 0708.2252).

6. Extensions: Polarization Strategies and Embeddings

Focused proof systems admit flexible polarization strategies—any assignment of atomic polarities yields a correct but distinct focused calculus. “Delay operators” (inserting trivial connectives to artificially control phase transitions) can be used to embed various proof search strategies (forward-chaining, backward-chaining, or hybrid) within a uniform focused framework. This enables recovery of classical systems such as LJT, LJQ, and $\Gamma \Rightarrow P$6RCC as subsystems of LJF through appropriate polarity and delay choices (0708.2252).

7. Focused Proofs in Bunched and Contextual Logics

In substructural logics such as BI, the combination of nested sequents, polarization, and focusing admits an analytic (cut-free and complete) goal-directed calculus (Gheorghiu et al., 2020). For contextual systems such as LF-based multi-level logics (e.g., Beluga), focusing can be applied to both specification and meta-logics, supporting uniform and focused phases at both levels. Automation strategies in such systems typically invoke maximal invertible decompositions, followed by focused handling of non-invertible connectives, which correspond directly to “interesting” case splits in meta-theoretic proofs (Schwartzentruber et al., 2023).

---

Focused proof systems, with their canonical phase structure, serve both as a minimal search space for sequent calculus proofs and as a foundation for mechanized reasoning in both foundational proof theory and applied (semi-)automated proof search. Their formal properties—soundness, completeness, normalization, and minimality—are now well understood in a range of logical settings from classical to substructural and dependently typed meta-logics (Gheorghiu et al., 2020, Simmons, 2011, Nigam et al., 2015, Schwartzentruber et al., 2023, 0708.2252).