e-Boost: Multi-Domain Enhancement Frameworks

• e-Boost in e-graph extraction uses parallel heuristic extraction, adaptive search space pruning, and warm-started ILP to achieve up to 558× speedup and notable area improvements in logic synthesis.
• In plasma and laser-based acceleration, e-Boost integrates compact energy booster stages that enable multi-GeV gains while preserving beam quality and reducing accelerator dimensions.
• e-Boost in statistical testing applies calibrated boost factors to e-value-based methods, enhancing power by 30–80% and lowering sample size requirements via adaptive thresholds.

The term "e-boost" is used across several technical fields to denote diverse mechanisms or frameworks that enhance underlying system performance via boosting, reweighting, or energy injection strategies. This article catalogs and analyzes prominent usages of "e-boost" spanning e-graph extraction algorithms, particle/ion acceleration, solid-state propulsion, and statistical inference, focusing on articles with the term in their title or its explicit adoption in recent literature.

1. e-Boost in E-Graph Extraction and Optimization

In logic synthesis and formal verification, "e-boost" refers to a high-performance framework for e-graph extraction—a core combinatorial optimization step wherein one selects optimal representatives for equivalence classes (e-classes) in an e-graph encoding exponentially many possible circuit rewritings or proof steps. E-graph extraction under realistic DAG (Directed Acyclic Graph) cost is NP-hard due to the requirement to count shared subexpressions only once.

The e-boost framework (Yin et al., 18 Aug 2025) introduces three technical innovations that address the trade-off between speed and optimality in previous approaches:

1. Parallelized Heuristic Extraction: Exploits weak data dependence in the standard greedy bottom-up solver to enable parallel computation. The per-node costs for all nodes in a batch can be updated in parallel, only requiring synchronization when updating per-class minimal cost sets. Pseudocode demonstrates batch processing, local update aggregation, and minimal lock contention.
2. Adaptive Search Space Pruning: Following the heuristic pass, nodes whose cost exceeds a multiplicative threshold $\theta \cdot \text{cost}_{\min}(c)$ are pruned per class. Typical choices are $\theta\in\{1.05,1.25,1.5\}$, reducing ILP instance sizes by multiple orders of magnitude without materially sacrificing the global optimum.
3. Initialized Exact Solving (Warm-Started ILP): An integer linear program (ILP) is formulated on the pruned subgraph. The ILP incorporates constraints enforcing per-class choice, DAG acyclicity, and explicit warm-start using the heuristic solution. This significantly shortens time to optimality.

In formal benchmarks, e-boost achieved up to 558$\times$ speedup over vanilla ILP solvers and outperformed the prior differentiable state-of-the-art (SmoothE) by 19.04% in solution cost, with area improvements in logic synthesis pipelines of 7.6–8.1% over baselines. Parallel speedups for the heuristic phase exceeded 20$\times$ on large e-graphs.

A summary of the workflow is presented below:

Stage	Algorithmic Strategy	Impact
Heuristic Extraction	Parallel, bottom-up	Drastic speedup
Search Space Pruning	Threshold $\theta$ cut	ILP size reduction
Exact Solving (ILP)	Warm-start, pruned vars	High solution quality

The framework is robust to batch scheduling, with parallel convergence guarantees matching sequential minima. In rare cases, pathological e-graphs may see optimal solutions pruned, but convergence speed gains are generally dominant (Yin et al., 18 Aug 2025).

2. e-Boost in Plasma-Based Particle Acceleration

In accelerator physics, "e-boost" designates a plasma-based energy booster stage, especially as applied to free-electron lasers (FELs) (Schroeder et al., 2024). Embedding a compact plasma wakefield accelerator (PWFA) after the main RF linac permits major energy upgrades without extending accelerator length.

Key technical features:

• Plasma Wakefield Acceleration: A drive electron bunch excites a nonlinear plasma wake; a trailing (witness) bunch samples the wake, gaining energy $\Delta E = e E_z L_b$.
• Booster Staging: The e-boost stage occupies $0.5$–$2$ meters, with plasma density $n_0 \approx 10^{16}$–$10^{18}\,\text{cm}^{-3}$, achieving $\theta\in\{1.05,1.25,1.5\}$0–$\theta\in\{1.05,1.25,1.5\}$1 GV/m, orders of magnitude above RF linac fields.
• Beam Quality Preservation: Using controlled current profiles and matching optics, the setup preserves emittance and yields sub-percent energy spread at GeV-level gains.

Simulations (HiPACE++/SALAME) validate resultant energy gains up to $\theta\in\{1.05,1.25,1.5\}$2 GeV at EuXFEL and LCLS-II scale facilities, with practical gains in FEL brilliance, reduced saturation length, and extended tunable range. Integration entails precise two-bunch generation, micron-scale focusing, and fs-level synchronization.

Facility	Pre-Boost Energy	e-Boost Gain	Outcome
LCLS-II	4 GeV	$\theta\in\{1.05,1.25,1.5\}$34 GeV	Sub-Å X-ray access
FLASH	1 GeV	$\theta\in\{1.05,1.25,1.5\}$4 GeV	Water window at high rep
EuXFEL	17.5 GeV	up to 14 GeV	Dual-energy beamlines

Current R&D focuses on stability under high-average-power operation and real-time active alignment (Schroeder et al., 2024).

3. Dual-Stage Laser e-Boost for Multi-GeV Electrons

Dual-stage laser-wakefield accelerators (LWFA) employ laser-driven plasma stages for "boosted" electron acceleration into the multi-GeV regime (Kim et al., 2013). The scheme employs:

• Injector: Short, high-density ($\theta\in\{1.05,1.25,1.5\}$5) helium jet provides self-injected, $\theta\in\{1.05,1.25,1.5\}$6400 MeV seed electron bunch.
• Accelerator: Longer, lower-density ($\theta\in\{1.05,1.25,1.5\}$7) plasma extends the dephasing and depletion lengths, enabling multi-GeV energy gain.

Single 1-PW, 30-fs laser pulses (normalized vector potential $\theta\in\{1.05,1.25,1.5\}$8) produce high-field bubbles with ultrafast injection. Experimental spectra reach $\theta\in\{1.05,1.25,1.5\}$9 GeV, as corroborated by 3D PIC simulations.

Scaling laws (Lu et al.) confirm energy gain maximization at lower plasma density and longer acceleration lengths. The dual-stage approach decouples injection and acceleration, overcoming density/length trade-offs present in single-stage LWFAs.

4. e-Boost in Electrohydrodynamic (EHD) Propulsion

In atmospheric propulsion research, "e-boost" refers to rotary electrohydrodynamic (EHD) self-boosted propellers that use ionic wind for lift (Ieta et al., 2019). These devices achieve thrust via high-field ionization at blade-edge electrodes, driving ion drift and momentum transfer to neutral air.

Key attributes of e-boost EHD rotors:

• Voltage Range: Liftoff achieved at $\times$0 kV (3.5 cm, 0.2 g rotor) up to $\times$1 kV (25.5 cm, 27.8 g).
• Thrust-to-Power Efficiency: Reaches $\times$2–$\times$3 N/kW (vs. $\times$42.5 N/kW for turbojets).
• Scaling: Thrust increases quadratically with applied voltage, supporting rotor scaling by two orders of magnitude without prohibitive voltage escalation.

Benefits include quiet, solid-state operation and high efficiency at both mass extremes. Drawbacks are the need for external high-voltage supplies and current thrust density limitations relative to motorized rotors. Further research targets integrated power electronics, electrode optimization, and CFD-driven design (Ieta et al., 2019).

5. e-Boosting in Multiple Testing via e-Values

In modern statistical inference, "e-boost" refers to procedures that systematically increase the power of e-value-based hypothesis tests and FDR-controlling algorithms by leveraging distributional shape assumptions (Blier-Wong et al., 2024).

Key results in statistical e-boosting:

• Improved Thresholds: The default rejection threshold $\times$5 for control of Type I error can be reduced if e-value densities are decreasing or unimodal, yielding thresholds $\times$6 or even $\times$7 when log-e-values are decreasing or symmetric.
• Supremum Construction for Comonotonic e-Values: If null e-values are comonotonic, using $\times$8 as a test statistic preserves size at lower thresholds.
• Boosting in e-BH: When the e-BH (e-value Benjamini–Hochberg) procedure is applied, e-values can be multiplied by a boost factor $\times$9 calibrated to the null distribution's shape, thus increasing power while maintaining false discovery control.

Empirical simulations show power gains of 30–80% and substantial reductions in required $\times$0 for the same power, contingent on histogram-based diagnostics of e-value shape. The practical application requires only simple monotonicity checks and root-finding for $\times$1, making e-boost attractive for large-scale inference workflows (Blier-Wong et al., 2024).

6. e-Boosted Ion Acceleration for Fusion and Diagnostics

In laser-cluster plasma physics, "e-boost" (quasi-dc pulse boosting) describes the method of accelerating ions from a laser-initiated Coulomb explosion by applying a high-voltage quasi-dc pulse. This achieves two goals: complete electron evacuation (maximizing ion energy) and deterministic acceleration of all cluster ions toward a dense solid-state fusion target (Kaplan, 2015).

• Mechanism: A laser pulse causes a bare Coulomb explosion; a quasi-dc field ($\times$2–$\times$3 kV across $\times$4 cm) accelerates ions, raising impact energies ($\times$5 from several keV up to $\times$6).
• Fusion Yield Optimization: The approach increases D$\times$7+D collision cross-section by boosting ion energies into the regime of maximal fusion probability, with further stacking via multi-layered targets.
• Diagnostics: Time-of-flight measurements on segmented cathodes enable reconstruction of initial ion energy spectra with sub-ns time resolution.

Bulk neutron yield per cluster remains $\times$8 due to underlying probabilities; thus, the system is best used for diagnostics and as a pulsed neutron source rather than fusion power production (Kaplan, 2015).

7. Comparative Perspective and Thematic Synthesis

The underlying principle uniting these disparate e-boost usages is the strategic injection of computational, energetic, or statistical "boost" to a process bottlenecked by resource, complexity, or physical constraints. Whether parallelism and ILP warm-start in e-graph extraction, plasma field enhancement in accelerators, statically-shaped boosting in hypothesis testing, or synchronized field application in plasma physics, e-boost consistently denotes a compound approach that produces pronounced gains in efficiency, power, or scalability.

Each domain leverages domain-specific theory for correctness and quantifies gains in concrete metrics—runtime acceleration, energy reach, FDR control, or fusion neutron yield. Common open challenges include calibration of boost parameters (thresholds $\times$9, voltage $\theta$0, boost factor $\theta$1), and robust design against pathological input distributions or edge-case system responses. The generalizable "e-boost" paradigm thus encapsulates adaptive, multi-level enhancement strategies across computational, physical, and statistical disciplines, with each instantiation continuing to evolve in line with technical advances and benchmarking best practices (Yin et al., 18 Aug 2025, Schroeder et al., 2024, Kim et al., 2013, Kaplan, 2015, Ieta et al., 2019, Blier-Wong et al., 2024).