TSC^i_1: Polylog-Space Bounded Arithmetic

• TSC^i_1 is a bounded arithmetic framework in L₁ that employs dependent-choice sequences to delineate fragments with exact poly-time, polylog-space (SC) computability.
• It is positioned between the logspace theory TLS^i_1 and the induction-based fragment 𝑆̂^i_1, illustrating refined calibration of bounded arithmetic hierarchies.
• The theory harnesses iterative schemata and precise growth bounds to encapsulate resource-limited computations and demonstrate nontrivial independence results, including MRDP separation.

$TSC^i_1$ is an arithmetic introduced in “Arithmetics within the Linear Time Hierarchy” (Pollett, 15 Aug 2025). It is defined in the language $L_1$ using axioms related to dependent choice sequences for formulas from two syntactic classes within $\Sigma^{\mathsf b}_i$, and it is part of a family of theories that also includes $\breve{S}^i_1$ and $TLS^i_1$ (Pollett, 15 Aug 2025). The theory is designed to isolate fragments of $S_1$ with strong closure properties and exact characterizations of their definable multifunctions; for $i=1$, its multifunction class simplifies to $SC$, Steve’s Class, namely poly-time, polylog-space computation (Pollett, 15 Aug 2025).

1. Formal setting in $L_1$

The theory is formulated in the language $L_1$ with non-logical symbols

$L_1$0

with intended meanings including $L_1$1, $L_1$2, $L_1$3, and $L_1$4 (Pollett, 15 Aug 2025). The same framework also defines coding operations such as $L_1$5, block projections $L_1$6, bit extraction $L_1$7, pairing, and tuple projections, so that computations and witness sequences can be encoded inside bounded arithmetic (Pollett, 15 Aug 2025).

The syntactic point of departure is the usual hierarchy of bounded formulas $L_1$8, where sharply bounded quantifiers are ignored when determining quantifier alternations (Pollett, 15 Aug 2025). The paper then introduces new classes by counting bounded existential and sharply bounded universal quantifier blocks. In this setting, bounded quantifiers are of the form $L_1$9 and $\Sigma^{\mathsf b}_i$0, while sharply bounded quantifiers are of the form $\Sigma^{\mathsf b}_i$1 and $\Sigma^{\mathsf b}_i$2 (Pollett, 15 Aug 2025). This refinement is used to build the iterative hierarchies that underlie $\Sigma^{\mathsf b}_i$3.

A central technical ingredient is a family of $\Sigma^{\mathsf b}_i$4-steppable and $\Sigma^{\mathsf b}_i$5-iterable formulas, which formalize bounded configuration updates and bounded dependent-choice sequences. The growth terms are chosen to track sublinear resources; in particular, the paper defines $\Sigma^{\mathsf b}_i$6 as an approximation to $\Sigma^{\mathsf b}_i$7, where $\Sigma^{\mathsf b}_i$8 is the “length of the length” (Pollett, 15 Aug 2025). This is the mechanism by which the formalism connects arithmetic definitions to time-space tradeoffs.

2. Definition of $\Sigma^{\mathsf b}_i$9

The theory is built over $\breve{S}^i_1$0, where

$\breve{S}^i_1$1

and $\breve{S}^i_1$2 is the finite quantifier-free axiom base for the language $\breve{S}^i_1$3 (Pollett, 15 Aug 2025). The defining scheme of $\breve{S}^i_1$4 is a restricted iteration principle:

$\breve{S}^i_1$5

Specialized to $\breve{S}^i_1$6, this gives

$\breve{S}^i_1$7

(Pollett, 15 Aug 2025).

The $\breve{S}^i_1$8 axioms assert existence for bounded iteration formulas encoding dependent choice sequences. In normalized form, the schema is

$\breve{S}^i_1$9

with $TLS^i_1$0 drawn from $TLS^i_1$1 and $TLS^i_1$2 in the relevant lower-level syntactic class (Pollett, 15 Aug 2025). Operationally, the schema says that if a step relation is sufficiently well-formed, then a bounded encoded computation sequence exists.

The syntactic classes feeding the definition are organized through the iterative closures $TLS^i_1$3, $TLS^i_1$4, and $TLS^i_1$5. In the notation given in the paper,

$TLS^i_1$6

with corresponding $TLS^i_1$7-restricted variants $TLS^i_1$8, $TLS^i_1$9, and $S_1$0 (Pollett, 15 Aug 2025). These classes formalize increasingly expressive families of iterable and query-iterable formulas.

The contrast with the companion theory $S_1$1 is exact. $S_1$2 uses the same general architecture but replaces the growth bound $S_1$3 by $S_1$4: $S_1$5 A plausible implication is that the difference between $S_1$6 and $S_1$7 is entirely concentrated in the size of the allowed dependent-choice encodings (Pollett, 15 Aug 2025).

3. Proof-theoretic position

For $S_1$8, the paper proves the chain

$S_1$9

(Pollett, 15 Aug 2025). This places $i=1$0 strictly between the logspace-oriented $i=1$1 and the induction-based fragment $i=1$2.

The intermediate theory $i=1$3 is defined as

$i=1$4

that is, a length-induction theory over the $i=1$5 class (Pollett, 15 Aug 2025). Its role is to mediate between direct induction on iterative formulas and existence axioms for dependent-choice sequences.

The conservativity relation

$i=1$6

is a universal $i=1$7-conservativity statement: if $i=1$8 proves a universal closure of a boolean combination of $i=1$9-formulas, then $SC$0 already proves it (Pollett, 15 Aug 2025). This gives the hierarchy a precise proof-theoretic calibration rather than merely a semantic one.

A notable feature of the framework is that it is developed in $SC$1, where the usual $SC$2-style quantifier-exchange mechanisms are unavailable. The paper explicitly motivates the new syntactic classes and the $SC$3 schemes as substitutes for stronger replacement-style principles that fail in this setting (Pollett, 15 Aug 2025). This suggests that $SC$4 is not a simple translation of $SC$5-style bounded arithmetic into a weaker language, but a tailored arithmetic for linear-time-hierarchy phenomena in $SC$6.

4. Definable multifunctions and computational meaning

The central characterization theorem states that the $SC$7-definable multifunctions in $SC$8 are exactly

$SC$9

(Pollett, 15 Aug 2025). The paper defines this as the class of $L_1$0-computable multifunctions, with output bounded by a term in $L_1$1, and with access to a witness oracle for the formula class $L_1$2 (Pollett, 15 Aug 2025).

For $L_1$3, the characterization simplifies: the multifunctions definable in $L_1$4 are precisely the functions in $L_1$5, Steve’s Class, meaning poly-time and polylog-space computable functions (Pollett, 15 Aug 2025). This is one of the paper’s main exactness claims and is the reason the notation $L_1$6 is tied to the $L_1$7 complexity class.

The companion result for $L_1$8 is the logspace analogue: $L_1$9 (Pollett, 15 Aug 2025). The pair $L_1$0 therefore separates logspace and $L_1$1 at the level of definability while preserving the same underlying syntactic strategy.

The paper also emphasizes closure properties. The definable multifunction classes admit composition, and the framework supports bounded search and witness extraction through the $L_1$2 and $L_1$3 operators (Pollett, 15 Aug 2025). A plausible implication is that the arithmetics were designed not only to represent single resource-bounded computations, but also to remain stable under the usual algebra of function construction.

5. Relation to the Linear Time Hierarchy

The stated goal is to identify fragments of $L_1$4 “within the Linear Time Hierarchy,” and the resource interpretation is made explicit through time-space simulation lemmas (Pollett, 15 Aug 2025). A key technical statement is a Nepomnjascii-type inclusion: $L_1$5 from which the paper concludes that $L_1$6 and $L_1$7 are contained in low levels of the $L_1$8–$L_1$9 iterative hierarchy over open formulas (Pollett, 15 Aug 2025).

The role of the growth bounds is decisive. Terms of size $L_1$00 encode configurations, while $L_1$01 controls the number of iteration steps (Pollett, 15 Aug 2025). In $L_1$02, the relevant bounds are polynomial in $L_1$03; in $L_1$04, they are of the form $L_1$05, which is the appropriate scale for polylog-space computations (Pollett, 15 Aug 2025).

The paper’s motivation is also comparative. In $L_1$06, theories such as $L_1$07 benefit from $L_1$08-based quantifier exchange and stronger collection behavior. In $L_1$09, Parikh-type obstacles prevent a direct transfer of those methods (Pollett, 15 Aug 2025). $L_1$10 is therefore part of a different program: it recovers exact computational meaning in a weaker language by restricting the syntactic form of dependent-choice sequences rather than enlarging the language.

Within that program, $L_1$11 plays the $L_1$12-side role. The paper’s abstract states that the $L_1$13-definable multifunctions are the $L_1$14 computable multifunctions whose output is bounded by a term in $L_1$15 and that have access to a witness oracle for $L_1$16 (Pollett, 15 Aug 2025). This places $L_1$17 squarely at the interface between bounded arithmetic and polylog-space complexity.

6. Independence results, examples, and significance

The paper proves that

$L_1$18

(Pollett, 15 Aug 2025). This is presented as an independence result related to the Matiyasevich–Robinson–Davis–Putnam theorem. The stated proof strategy is that if $L_1$19 proved MRDP, then one would derive collapses such as $L_1$20 and ultimately $L_1$21, contradicting the separation arguments used in the paper (Pollett, 15 Aug 2025).

A second family of independence results concerns whether the theories prove that simultaneous nondeterministic polynomial time, sublinear space is equal to co-nondeterministic polynomial time, sublinear space (Pollett, 15 Aug 2025). The paper states that $L_1$22 cannot prove the corresponding equality between the existential and universal $L_1$23–$L_1$24 hierarchy levels (Pollett, 15 Aug 2025). This situates $L_1$25 as strong enough to capture $L_1$26, but not so strong as to collapse the sublinear-space alternation structure it is meant to study.

The theory is also illustrated by explicit definability constructions. One example is $L_1$27, the number of $L_1$28-bits in $L_1$29, which the paper shows is definable already in $L_1$30 via nested iteration over encoded blocks (Pollett, 15 Aug 2025). More generally, the framework provides formulas such as $L_1$31 and $L_1$32 for simulating oracle Turing-machine computations inside bounded arithmetic (Pollett, 15 Aug 2025). These examples are significant because they demonstrate that the iterative axioms are not merely proof-theoretic devices; they directly encode concrete machine computations.

The broader significance of $L_1$33 lies in the combination of three properties established in the paper: syntactic precision, exact complexity-theoretic characterization, and nontrivial independence behavior (Pollett, 15 Aug 2025). In that sense, $L_1$34 is a resource-calibrated arithmetic tailored to $L_1$35 and to the linear-time-hierarchy viewpoint, rather than a generic bounded arithmetic fragment.