Sequence Lock: Domain-Specific Mechanisms

• Sequence lock is a term for domain-specific mechanisms that regulate permissible state changes through ordered sequences.
• In computational geometry and quantum sensing, sequence locks enable configurations adjustments and pulse-based detection to overcome inherent constraints.
• In hardware security and databases, sequence locks secure scan operations and serialize transactions to improve system robustness.

Sequence lock is not a single standardized technical notion. In current research literatures, the expression and closely related forms denote several distinct constructs: the geometric question of whether a polygonal chain can lock under continuous non-self-intersecting motion; pulse-sequence methods such as spin-lock induced crossing for singlet preparation in NMR; quantum lock-in pulse trains that modulate phase accumulation in trapped-ion and many-body interferometers; secure scan-locking of sequential circuits; and queue- or group-based lock sequencing in database concurrency control [9908005, (DeVience et al., 2013, Shaniv et al., 2016, Zhuang et al., 2020, Potluri et al., 2020, Wang et al., 9 Apr 2025)]. This suggests that “sequence lock” functions less as a universal term than as a field-specific label for mechanisms in which admissible state changes are controlled by an ordered sequence.

1. Terminological scope

Across the cited work, the word “lock” refers to different technical objects: geometric entrapment, RF-mediated state transfer, narrow-band quantum demodulation, scan-path obfuscation, and transaction serialization. The common ingredient is sequencing, but the controlled entity differs sharply by domain.

Domain	Meaning of lock	Sequencing element
Computational geometry	A chain or tree cannot reach a canonical configuration	Continuous reconfiguration path
NMR spectroscopy	Spin-lock induced crossing between singlet and triplet levels	RF pulse sequence and spin-lock duration
Quantum sensing	Quantum lock-in detection of an AC signal	Periodic train of equidistant $\pi$-pulses
Hardware security	Secure scan-locking of sequential circuits	FI-SQ locking and scan-chain behavior
Database systems	Ordered contention management under strict 2PL	Queue locking and group locking

Two explicit terminological cautions recur. The sequential-circuit paper states that its subject is not software sequence locks or database synchronization primitives, but scan-locking of sequential logic (Potluri et al., 2020). The TXSQL paper is likewise explicit that it is not about classic OS-style seqlocks, but about restructuring the order in which transactions queue, acquire locks, execute hotspot updates, commit, and roll back under high contention (Wang et al., 9 Apr 2025).

2. Geometric locking of polygonal chains

In computational geometry, locking concerns whether a polygonal linkage can be continuously reconfigured while preserving edge lengths and non-self-intersection. A polygonal chain is a sequence of rigid line segments connected end-to-end by joints; an open chain has two endpoints of degree 1, a closed chain is a polygonal cycle, and a tree generalizes an open chain by allowing branching. An open chain is straightened when all edges become collinear in order, whereas a closed chain is convexified when it reaches a convex polygonal configuration.

The central high-dimensional result is that, in all dimensions $d \ge 4$, every simple open polygonal chain and every tree may be straightened, and every simple closed polygonal chain may be convexified [9908005]. The abstract further states that these reconfigurations can be achieved by algorithms that use polynomial time in the number of vertices and result in a polynomial number of moves. The theorem is dimension-sensitive: it contrasts with $d=2$, where trees can lock, and with $d=3$, where open and closed chains can lock [9908005].

The significance of this result lies in the non-monotone role of ambient dimension. The planar setting is governed by carpenter’s-rule-type nonlocking for simple open chains and convexification for simple polygons, yet the three-dimensional setting admits genuinely locked open chains, closed chains, and trees. The four-dimensional theorem shows that the additional degree of freedom is sufficient to eliminate those obstructions. A plausible implication is that, in this literature, “lock” is best understood as a property of the admissible configuration space rather than of local kinematic constraints.

3. Spin-lock induced crossing in NMR

In NMR, one important sequence-based meaning of lock arises in spin-lock induced crossing (SLIC), a pulse-sequence method for preparing nuclear spin singlet states in nearly equivalent spin-$1/2$ pairs (DeVience et al., 2013). The motivation is that, for two strongly $J$-coupled, nearly equivalent nuclei, the singlet state can have a lifetime much longer than the usual spin-lattice relaxation time $T_1$. Direct RF excitation does not efficiently transfer population between triplet and singlet manifolds because they have different exchange symmetry, and in an ideal equivalent-spin pair the singlet is symmetry-protected.

SLIC addresses this by applying a continuous resonant RF spin-lock with nutation frequency matched to the scalar coupling, $\nu_n = J$, so that in the rotating frame the energy of one triplet state coincides with that of the singlet (DeVience et al., 2013). At that crossing, the small symmetry-breaking term associated with the resonance frequency difference $\Delta \nu$ mixes the states and drives triplet-singlet polarization transfer. The paper gives the central timing rule for maximum transfer as

$t_{SL,\max} = \frac{0.707}{\Delta \nu}.$

The complete experiment consists of a $d \ge 4$0 preparation pulse creating transverse triplet polarization, a spin-lock at the SLIC condition, a singlet evolution interval $d \ge 4$1, a second identical spin-lock for readout, and inductive detection of the recovered transverse magnetization.

The paper compares SLIC to the older M2S sequence. M2S generates singlet population only during the final one-third of the preparation time, so more polarization can be lost to ordinary relaxation before reaching the long-lived singlet state. The ideal M2S maximum-transfer time is reported as

$d \ge 4$2

whereas SLIC reaches maximum singlet transfer about 40% faster (DeVience et al., 2013). The advantage is especially pronounced when $d \ge 4$3.

Experimentally, the method was demonstrated at 4.7 T on a 20 mM solution of phenylalanine-glycine-glycine in $d \ge 4$4, using two nearly equivalent proton pairs. For the $d \ge 4$5 ppm pair, the paper reports $d \ge 4$6 Hz from SLIC, $d \ge 4$7 Hz, $d \ge 4$8 ms, and $d \ge 4$9 s. For the $d=2$0 ppm pair, it reports $d=2$1 Hz, $d=2$2 Hz, $d=2$3 ms, and $d=2$4 s (DeVience et al., 2013). The round-trip triplet $d=2$5 singlet $d=2$6 triplet transfer was about 34% for the first pair and about 12% for the second, compared with 24% and 4% for M2S on the same systems. Because the theoretical maximum is 50%, the reported efficiencies imply 82% efficiency for each SLIC application on the first pair and 49% on the second (DeVience et al., 2013).

Within this usage, “lock” refers neither to mutual exclusion nor to geometric immobility. It denotes a continuous spin-locking field that engineers a rotating-frame degeneracy and thereby enables otherwise symmetry-forbidden population transfer.

4. Quantum lock-in sequences in precision sensing

A second sequence-based usage appears in quantum sensing, where a lock-in sequence is a Ramsey-type interferometric protocol with a periodic train of $d=2$7-pulses inserted during interrogation. In a single trapped $d=2$8 ion, the sequence is

$d=2$9

with total interrogation time $d=3$0 (Shaniv et al., 2016). The ion is electrically driven at about $d=3$1, far below the axial trap resonance $d=3$2, so the resulting displacement is tiny. Rather than measuring displacement directly, the experiment reads out the velocity-induced Doppler shift on the optical clock transition and uses the pulse train to make phase contributions from successive force half-cycles add coherently.

In toggling-frame language, the detuning is multiplied by a square-wave modulation function $d=3$3 that alternates between $d=3$4 and $d=3$5 (Shaniv et al., 2016). For the synchronized case, the paper gives the matching choice

$d=3$6

so that the sign flips of the sequence align with the force half-cycles. This converts an oscillating Doppler detuning into a monotonically accumulated phase while rejecting static and low-frequency detuning noise. The reported synchronized force sensitivity is as low as $d=3$7, with a measured oscillation amplitude $d=3$8 corresponding to $d=3$9 (Shaniv et al., 2016). In the asynchronous regime, where the initial phase is effectively uniformly sampled, the mean phase averages away and information is encoded instead in Bessel-function contrast decay; the reported sensitivities are $1/2$0 for $1/2$1 and $1/2$2 for $1/2$3 (Shaniv et al., 2016).

The many-body extension formulates the same idea as a quantum lock-in amplifier implemented within many-body quantum interferometry (Zhuang et al., 2020). The probe is an ensemble of $1/2$4 identical two-state bosonic particles described by collective-spin operators $1/2$5. During interrogation, a periodic train of sharp $1/2$6-pulses,

$1/2$7

defines a reference frequency

$1/2$8

and generates a square-wave toggling function that mixes the unknown AC field down to the beat frequency $1/2$9 (Zhuang et al., 2020). Under $J$0 and $J$1, the effective interaction is approximated by

$J$2

This framework allows both the frequency and amplitude of the unknown alternating field to be extracted from population measurements. For spin coherent states, the derived precisions are SQL-limited, scaling as $J$3. For GHZ and spin cat states, combined with interaction-based readout, the paper derives Heisenberg-like scaling, with both frequency and amplitude precisions approaching $J$4 (Zhuang et al., 2020). The paper also states that the amplifier remains robust against extreme stochastic noise.

In both the single-ion and many-body cases, the “lock” is a temporal reference waveform embedded in the control sequence. It is therefore a quantum analogue of classical narrow-band lock-in detection, not a concurrency-control primitive.

5. Sequential scan-locking in hardware security

In hardware security, “sequence lock” refers to sequential logic locking, specifically scan-locking of sequential circuits. The attack surface is the scan chain: scan infrastructure used for manufacturing test can be abused to initialize flip-flops, apply capture cycles, and observe internal responses, thereby reducing a sequential decryption problem to a combinational-style attack problem (Potluri et al., 2020). The SeqL scheme is designed for a malicious foundry or outsourced fabrication-and-test adversary with reverse-engineered gate-level access and scan access in EDT-bypass mode.

The central design principle is functional isolation. SeqL distinguishes the functional output $J$5 or $J$6 from the scan output $J$7, and locks selective flip-flop functional-input/scan-output pairs, or FI-SQ pairs (Potluri et al., 2020). The sequential key is decomposed as

$J$8

where $J$9 and $T_1$0 affect normal functional behavior and $T_1$1 affects only scan behavior. Because the scan path is locked independently of the functional path, an attacker can recover a key that is correct for scan observations yet incorrect for real functionality.

The formal model is expressed through scan-path parity and the Key Assignment Graph (KAG), a vertex-labeled, edge-weighted directed graph. The paper proves that KAG is a binary tree and derives the main security expression: if $T_1$2 FI-SQ pairs are locked, the odds against the functionally correct key among the scan-correct keys is

$T_1$3

For the illustrative two-pair example, there are four scan-correct keys and only one functionally correct key, so $T_1$4 (Potluri et al., 2020).

SeqL also includes automation. For pipelined combinational circuits the paper proposes the Iterative Key Pushing Algorithm (IKPA); for sequential circuits it proposes the Iterative Boundary Locking Algorithm (IBLA). Both incrementally lock FI-SQ boundaries, select flip-flops without feedback, and use SAT-guided iteration until functional corruption is achieved or a budget is exhausted (Potluri et al., 2020).

The reported empirical results are extensive. SeqL was evaluated on ISCAS and MCNC pipelined combinational benchmarks, ITC’99 sequential benchmarks, and a flattened RISC-V CPU. The paper states that SeqL gave 100% resilience to SAT, Double-DIP, HackTest, SMT, FALL, Shift-and-Leak, and Multi-cycle attacks (Potluri et al., 2020). On the sequential benchmarks and the RISC-V design, EFF resilience is reported as 0% while SeqL resilience is reported as 100%. SeqL overheads in that table range from 0.01% to 0.24% on the ITC benchmarks and 0.09% on the flattened RISC-V CPU, whereas EFF overheads range from 3.2% to 7.9% (Potluri et al., 2020). At the flip-flop level, the SeqL cell uses 50 transistors, with $T_1$5 ps and EPT $T_1$6 fJ, compared with 48 transistors, $T_1$7 ps, and $T_1$8 fJ for EFF (Potluri et al., 2020).

Here, a sequence lock is not a temporal pulse train but a sequential-circuit obfuscation mechanism that exploits the difference between scan-visible correctness and true functional correctness.

6. Lock sequencing in database concurrency control

In database systems, TXSQL studies lock optimization under high-contention workloads, beginning from strict two-phase locking. Under strict 2PL, transactions acquire locks in the growing phase and release them only after commit, so conflicting transactions must wait for the holder to commit and release its locks (Wang et al., 9 Apr 2025). The paper argues that hotspot access exacerbates lock contention and waiting, causing severe throughput degradation.

TXSQL introduces four layered mechanisms. The first is lightweight lock management based on a lock-free hash design, with a structure named trx_lock_wait whose key is a record ID and whose value is a queue of waiting transactions (Wang et al., 9 Apr 2025). The second is a copy-free active transaction list using a transaction attribute del_ts so visibility can be determined from snapshot information, version information, and del_ts without copying or locking the active transaction list. The third is queue locking: once a hotspot row is detected, its unique row identifier is inserted into a hotspot hash table, and transactions updating that row queue before contending for the actual lock manager.

The fourth mechanism, hotspot-aware group locking, is the paper’s main sequencing contribution. A row is considered a hotspot when the number of waiting transactions exceeds a threshold, given as a rule of thumb of 32 (Wang et al., 9 Apr 2025). Transactions updating the same hotspot row are organized into a conflict group. One leader acquires the row lock; followers do not acquire row locks but execute hotspot updates serially in update order and in an uncommitted state. Each hotspot update receives a globally increasing identifier hot_update_order via global_hot_update_order.fetch_add(1), and the row maintains a dependency list (Wang et al., 9 Apr 2025).

Correctness is enforced by two explicit sequencing laws:

$T_1$9

and

$\nu_n = J$0

The leader does not release the lock until all already-granted followers have completed their hotspot updates, which prevents races during leader switching (Wang et al., 9 Apr 2025). The paper also reuses TRX_UNDO_TRX_NO in the undo log header so that, on restart, active transactions can be reordered by hot_update_order and unfinished hotspot transactions rolled back sequentially in order.

The reported performance gains are substantial under high contention. The abstract states that TXSQL achieves performance improvements of up to 6.5x compared to state-of-the-art methods and up to 22.3x compared to state-of-the-art systems (Wang et al., 9 Apr 2025). In FiT, the paper reports throughput improvements of up to $\nu_n = J$1 over MySQL, O1, and O2. In SysBench hotspot update, TXSQL achieves up to $\nu_n = J$2 improvement, and at 1024 threads the reported TPS values are 2627.1 for MySQL and 19958.37 for TXSQL (Wang et al., 9 Apr 2025). The paper also reports deployment to over 20,000 database instances, with over 30% performance improvement in non-hotspot scenarios and up to 10x in hotspot scenarios.

The limitations are explicit. The current solution supports only transactions updating a single hotspot row, cascading rollback can degrade performance sharply, and hotspot mechanisms do not help when updates are dispersed across multiple hotspots rather than concentrated on one severe hotspot (Wang et al., 9 Apr 2025). The paper further stresses that this is not a classic seqlock design with version counters and lock-free readers, but a database-specific strategy for ordered contention management.

7. Comparative interpretation

The surveyed usages demonstrate that “sequence lock” is a polysemous research term. In geometry, lock means an obstruction in configuration space; in NMR and quantum sensing, it denotes a pulse-sequence-mediated resonance or demodulation condition; in hardware security, it denotes scan-locking of sequential logic; and in databases, it denotes ordered queueing and serialization policies for conflicting transactions [9908005, (DeVience et al., 2013, Shaniv et al., 2016, Zhuang et al., 2020, Potluri et al., 2020, Wang et al., 9 Apr 2025)].

A common misconception is to treat these literatures as variants of one synchronization primitive. The papers themselves reject that simplification. SeqL is explicitly not about software locks or sequence locks in concurrent programming (Potluri et al., 2020), and TXSQL is explicitly not about classic OS-style seqlocks (Wang et al., 9 Apr 2025). Conversely, the quantum lock-in papers use “lock” in the metrological sense of a reference-modulated, frequency-selective accumulation protocol, and the polygonal-chain paper uses it in the geometric sense of reconfiguration impossibility.

This suggests that the most accurate encyclopedic treatment is comparative rather than unificatory. The term does not name a single formal object across fields. Instead, it marks a family of domain-specific mechanisms in which an ordered sequence—of configurations, RF controls, $\nu_n = J$3-pulses, scan interactions, or transaction events—determines whether motion, transfer, observation, or access is permitted.