Ice Crystal Regulation in Aquatic Products via Physical‐Field‐Assisted Freezing: Mechanisms and Research Progress

ABSTRACT

Aquatic products provide one‐fifth of global animal protein and are rich in long‐chain n‐3 polyunsaturated fatty acids. Their fragile muscle structure, high moisture, unsaturated lipids, and endogenous enzymes make them prone to quality loss during freezing. Traditional freezing forms large uneven ice crystals, causing 10%–25% drip loss and rapid lipid oxidation even under standard cold‐chain conditions. Various physical‐field‐assisted freezing methods can regulate supercooling, ice nucleation, and unfrozen water distribution to optimize ice crystal formation. This review analyzes freezing‐induced damage from microstructural damage, water migration, molecular instability, and species‐specific biochemistry and evaluates technologies by mechanism validity, applicable range, species adaptability, and practical operability. Final product quality is largely determined by the degree of supercooling at nucleation, the rate of latent‐heat removal during crystal growth, and the extent of field‐induced perturbation to native proteins and lipids. Excessive treatment intensity will trigger tissue deterioration. Currently, high‐pressure and continuous ultrasound‐assisted freezing have mature mechanisms and applicable parameters, suitable for high‐value aquatic products despite limited production capacity. By contrast, magnetic and low‐frequency electric field‐assisted freezing lack solid theoretical support, with unstable practical effects. This article classifies these techniques by mechanism, operation range, and engineering practicability and puts forward targeted application strategies and standardized research norms. Future studies should resolve contested mechanisms, test hybrid protocols under factorial designs, harmonize reporting standards, and validate scale‐up performance.

Abbreviations

ABF

air‐blast freezing

AF

air freezing

EFAF

electric field‐assisted freezing

F/T

freeze–thaw

FD

fractal dimension

HPAF

high‐pressure‐assisted freezing

HPT

high‐pressure thawing

IF

immersion freezing

IQF

individually quick frozen

LNF

liquid‐nitrogen freezing

LNQF

liquid‐nitrogen quick‐freezing

MFAF

magnetic field‐assisted freezing

MUIF

multifrequency ultrasound‐assisted immersion freezing

MWAF

microwave‐assisted freezing

PF

plate freezing

PSF

pressure‐shift freezing

RFAF

radiofrequency‐assisted freezing

UAF

ultrasound‐assisted freezing

UIF

ultrasound‐assisted immersion freezing

US‐HPAF

ultrasound‐assisted high‐pressure freezing

1. Introduction

Aquatic products are a globally vital source of high‐quality animal protein; long‐chain ω‐3 polyunsaturated fatty acids, including eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), fat‐soluble vitamins (A, D, and B12); and essential trace elements (selenium, zinc, and iron) (Calder 2020; Glencross et al. 2024; Makay et al. 2026). Global fisheries and aquaculture production reached a record 223.2 million tons in 2022. For the first time, aquaculture production exceeded capture fisheries, and apparent per‐capita consumption of aquatic animal foods reached 20.7 kg. International trade in aquatic products exceeded $195 billion, and the sector is projected to grow by a further 10% by 2032 (FAO 2024). Sustaining this transcontinental supply chain—spanning cod, salmon, tuna, shrimp, and crab—depends overwhelmingly on cold‐chain logistics, with more than half of all aquatic products preserved by freezing and frozen storage (Huang et al. 2024; Liu, Tan, et al. 2025).

Aquatic muscle is intrinsically perishable because of its high moisture content (70%–85%), loose myofibrillar architecture, abundant unsaturated lipids, and active endogenous enzymes, including proteases, lipases, and lipoxygenases (Russo et al. 2024; Stella et al. 2022). In addition, psychrotrophic bacteria, especially Pseudomonas, Shewanella, and Aeromonas spp., contribute to rapid postmortem softening, oxidation, and off‐odor formation (Nie et al. 2025). Freezing preservation remains the most effective countermeasure, suppressing microbial and enzymatic activity to enable long‐term storage and intercontinental distribution (Du et al. 2022; FAO 2024; He et al. 2024; Liu et al. 2025).

Conventional freezing methods, like air‐blast, plate, fluidized‐bed, and liquid‐nitrogen freezing (LNF), rely on enhancing macroscopic heat transfer to shorten residence within the zone of maximum ice–crystal formation (−1°C to −5°C) (Huang et al. 2024; Jin et al. 2025; Zhang et al. 2024). Nevertheless, the heterogeneous tissue architecture and nonuniform temperature fields generate large, unevenly distributed ice crystals that pierce membranes and disrupt myofibrils, producing 10%–25% drip loss and flavor leaching upon thawing (Huang et al. 2024; Subhashini et al. 2024). In addition, the ±1–2°C temperature fluctuations typical of cold‐chain transport and storage promote Ostwald ripening. The result is the cross‐sectional area of ice crystals roughly doubles within 3 months storage at −18°C, whereas drip loss climbs from 5%–10% at start to 15%–26%, accompanied by softening, protein denaturation, and lipid oxidation (Du et al. 2022; Hu and Xie 2021; Jia, Roy, et al. 2022; Liu et al. 2025; Mao et al. 2025; Tan et al. 2022; Tan et al. 2021a, 2021b).

These limitations have motivated interest in physical‐field‐assisted freezing (Figure 1), including high‐pressure‐assisted freezing (HPAF)/ pressure‐shift freezing (PSF), ultrasound‐assisted freezing (UAF), magnetic field‐assisted freezing (MFAF), electric field‐assisted freezing (EFAF), and radiofrequency‐assisted freezing (RFAF)/microwave‐assisted freezing (MWAF). However, the mechanistic evidence differs substantially among these approaches. HPAF/PSF is mainly supported by pressure‐induced depression of the ice I liquidus and rapid, spatially distributed nucleation after pressure release (Subhashini et al. 2024). UAF is generally attributed to acoustic cavitation, microstreaming, and secondary nucleation or crystal fragmentation (Yu and Xie 2023). By contrast, the proposed effects of MFAF and low‐field EFAF on water structure, hydrogen‐bond networks, or nucleation barriers remain debated and have not been consistently verified in aquatic muscle (Peng et al. 2025; Pu et al. 2025). RFAF/MWAF is better interpreted as pulsed dielectric regulation of the unfrozen interfacial phase and temperature oscillation near the freezing front, rather than as a universal direct nucleation‐control mechanism (Jiang et al. 2023; Wang et al. 2025; Zheng, Zhang, et al. 2024). Each technique has its own limitations as follows: HPAF denatures proteins above 200–300 MPa (Suárez‐Medina et al. 2024; Subhashini et al. 2024); UAF generates sonochemical radicals and lipid oxidation (Zheng, Zou, et al. 2024); MFAF and EFAF mechanisms remain contested with poor reproducibility (Peng et al. 2025; Yang, Yang, et al. 2024); and industrial‐scale, food‐grade equipment is still immature (Xia et al. 2026).

FIGURE 1.

Ice crystal engineering in aquatic cold chain: from conventional freezing to physical field‐assisted technologies. EFAF, electric field‐assisted freezing; HPAF, high‐pressure‐assisted freezing; MFAF, magnetic field‐assisted freezing; MWAF, microwave‐assisted freezing; RFAF, radiofrequency‐assisted freezing; UAF, ultrasound‐assisted freezing.

Existing reviews address several adjacent/related topics—ice–crystal formation (Zheng, Zhang, et al. 2024), cryoprotectants (Wu et al. 2019), LNF (Huang et al. 2024), PSF (Subhashini et al. 2024), and physical‐field‐assisted freezing of muscle foods generally (Jiang et al. 2023)—but none links aquatic‐muscle peculiarities to dual macro‐micro regulation across the full field spectrum, nor critically benchmark mechanism clarity, cross‐species robustness, and engineering feasibility. The present review addresses this by treating freezing sensitivity as species‐dependent (lean fish, fatty fish, crustaceans, and cephalopods), by examining the conditions under which the published results disagree, and by ranking the technologies in terms of engineering readiness. Section 2 examines aquatic muscle structure; Section 3, ice–crystal dynamics; Section 4, the five technologies; Section 5 presents the integrated comparative framework, and Section 6 closes with synthesis and outlook.

2. Muscle Characteristics of Aquatic Products and Freezing Sensitivity Mechanisms

The exceptional freezing sensitivity of aquatic products cannot be explained by a single factor. Instead, it results from the interaction of four linked factors as follows: microstructural fragility, moisture redistribution, molecular instability, and species‐specific biochemistry. These factors do not act independently. Ice–crystal growth first disrupts the weak muscle architecture, which then promotes water migration, protein denaturation, lipid oxidation, and species‐dependent quality loss. Therefore, this section is organized as a mechanistic ladder from tissue geometry to water distribution, molecular deterioration, and species‐specific failure modes. Each level also defines a boundary condition that physical‐field‐assisted freezing strategies must respect.

2.1. Structural Basis of Freezing Sensitivity: Microstructural Fragility of Muscle Tissue

Fish skeletal muscle is segmentally organized into short myomeres separated by thin perimysial sheets—the myocommata—that contain only 3% collagen, compared with 15% in mammalian muscle, with low hydroxyproline content and limited intermolecular cross‐linking (Schmidt and Mouritsen 2022). Each myomere comprises fibers rarely longer than 20 mm and 20–100 µm in diameter, with an endomysium thinner than that of mammalian longissimus (Liu et al. 2025; Zhu et al. 2025). During slow freezing, ice crystals preferentially nucleate in the perimysial space. As these ice crystals grow, they compress adjacent muscle fibers and prestress the myocommata. Upon thawing, this structural distortion contributes to the characteristic “gaping” defect and increases drip loss to 10%–25% (Du et al. 2022; Zhu et al. 2025). However, recent electron‐microscopy and proteomic data place the cytoskeleton—not the connective tissue alone—at the center of damage: Z‐disc disruption and α‐actinin release become detectable after only 3–8 freeze–thaw cycles in carp (Wan et al. 2023) and in Trachurus spp. (Hu and Xie 2021), well before macroscopic gaping appears.

Crustaceans and cephalopods deviate sharply from this phenomenon. Shrimp and crab fibers are short, highly hydrated, and supported by very little connective tissue, leaving them even more vulnerable to crystal‐induced rupture (Teng et al. 2025). Cephalopod mantle, by contrast, is dominated by circumferential fibers encased in densely cross‐linked Types I and V collagen, conferring partial cryoresistance at the cost of severe textural toughening when collagen aggregates around advancing ice fronts (Schmidt and Mouritsen 2022). Even at the level of muscle architecture, the differences between taxa are large enough that a single protocol is unlikely to be optimal across species; whether this rules out a universal protocol on mechanistic grounds is taken up in Section 4.7.

2.2. Amplification Effect of Moisture Distribution on Freezing Sensitivity

Aquatic muscle contains 70%–85% water, partitioned by low‐field NMR T ₂ relaxometry into three populations: bound water (T ₂₁, <10 ms), immobilized water (T ₂₂, 30–100 ms; >90% of total, retained within the myofibrillar lattice), and free water (T ₂₃, >100 ms; loosely held in extracellular spaces) (Zhu et al. 2025), as commonly defined in low‐field NMR of muscle, with cut‐offs varying slightly between studies. Their freezing points differ (free water nucleates near −1°C, bound and immobilized water near −5°C), so extracellular ice forms first, and the resulting cryo‐osmotic gradient extracts T ₂₂ water from the lattice. The kinetic signature is a decline in T ₂₂ amplitude, a lengthening of its relaxation time, and a rise in T ₂₃: Water moves from the “reservoir” to the “channel” and ultimately exits as drip (Tan et al. 2025).

This effect is amplified with respect to the red meat for two reasons. First, fish fibers carry a smaller fraction of bound water and a more porous filament lattice, so frozen–thawed tissue is irreversibly more permeable than its fresh counterpart (Liu et al. 2025; Russo et al. 2024). Second, the migrated water carries dissolved sarcoplasmic proteins, free amino acids, and nucleotides that determine flavor, so a given mass of drip removes disproportionately more sensory and nutritional value than in mammalian muscle (Liu et al. 2024). Temperature fluctuations of ±1–2°C around −18°C repeatedly melt small intracellular crystals and refreeze them onto larger extracellular ones, shifting water from the T ₂₂ to the T ₂₃ population at each cycle (Chen, Wang, Li, et al. 2025). Because this transfer is essentially irreversible, drip loss continues to rise during storage even without further structural injury.

2.3. Molecular Instability: Further Aggravation of Quality Deterioration

Three reinforcing biochemical pathways detailed below continue to operate at sub‐zero temperatures even when crystal damage is minimized.

Myofibrillar protein (MP) denaturation: Solute exclusion from the ice phase concentrates the unfrozen aqueous fraction by an order of magnitude, depressing local pH below 5.5 and elevating ionic strength to levels that destabilize myosin's α‐helix, expose hydrophobic surfaces, and drive disulfide‐mediated aggregation (Lee, Jo, et al. 2024). Very rapid freezing without cryoprotection produces a large ice–water interfacial area that itself drives protein unfolding at the interface, so the freezing rate cannot be optimized in isolation from the resulting interfacial area (Chen, Wang, Li, et al. 2025).

Lipid oxidation: Aquatic phospholipids are exceptionally rich in long‐chain n‐3 PUFA (polyunsaturated fatty acids), principally EPA and DHA, whose multiple bis‐allylic methylenes are highly susceptible to radical chain propagation catalyzed by released heme iron, free Fe²⁺, and lipoxygenases that retain measurable activity down to approximately −10°C in cellular environments (Suárez‐Medina et al. 2024). The resulting hydroperoxides decompose into aldehydes (hexanal, nonanal, and 2,4‐decadienal) responsible for the rancid off‐flavors of long‐stored fatty fish (Calder 2020; Stella et al. 2022).

Endogenous enzymes and lipid–protein cross‐talk: Cathepsins, calpains, collagenases, and trimethylamine N‐oxide (TMAO) demethylase remain catalytically competent at the frozen storage temperatures, and non‐heme iron in dark muscle accelerates the latter reaction (Li, Wang, Yanagita, et al. 2024). The three biochemical pathways are not independent: Secondary lipid‐oxidation products, such as malondialdehyde and 4‐hydroxynonenal, covalently modify myosin thiols and lysines, coupling lipid oxidation to protein aggregation in a positive feedback that low temperature alone does not interrupt (Bao et al. 2021). Thawing exudates collected from carp fillets frozen and stored for as little as 1 month already display elevated carbonyl and dityrosine contents and induce further oxidation when re‐incubated with MP (Liu et al. 2024), illustrating how damage continues to propagate after the freeze–thaw event itself.

2.4. Species‐Specific Differences in Freezing Sensitivity

To apply these mechanisms into actual production practice, we propose four operational categories whose dominant failure modes are qualitatively distinct.

Lean white‐fleshed fish, including cod, hake, Pollock, and sea bass, contain less than 2% total lipid, so lipid oxidation is seldom the main factor limiting their storage quality. Their most prominent quality issue is the spongy and fibrous tissue texture. This defect arises from formaldehyde accumulation. Such substance is produced by TMAO demethylase from endogenous TMAO at concentrations ranging from 60 to 140 mM, and this enzyme mainly exists in dark muscle, blood vessels, and internal organs (Li, Wang, Yanagita, et al. 2024). Bleeding quality and deboning practice are therefore at least as critical as the freezing process itself.

Fatty pelagic fish like salmon, mackerel, tuna, and sardines contain 10%–20% total fat, with polyunsaturated fatty acids accounting for over 30% of total lipids. Lipid oxidation becomes the primary factor shortening their shelf life. To effectively extend storage time, these fish need to be kept below −30°C, while being protected from light and oxygen exposure (Suárez‐Medina et al. 2024). Intramuscular lipids can ease ice‐induced damage to protein structures to some extent, yet they bring extra risks. Lipid‐derived free radicals speed up myoglobin oxidation, which further leads to the yellow discoloration of salmon muscle tissue (Russo et al. 2024).

For crustaceans, such as snow crab and crayfish, their thin outer shell and sparse connective tissues provide barely any mechanical protection against freezing injury. After thawing, polyphenol oxidase triggers rapid black discoloration within hours, and this quality defect often occurs alongside texture deterioration caused by serine protease activity (Tian et al. 2025). Hence, efficient freezing needs to combine fast heat removal with PPO inhibition methods. Liquid nitrogen freezing keeps crayfish drip loss under 8% even after five freeze–thaw cycles, whereas conventional AF at −20°C causes drip loss to rise above 20% under the same treatment conditions (Teng et al. 2025).

Squid, octopus, and other cephalopods possess tightly cross‐linked Types I and V collagen, along with circular muscle structure, which grants them certain natural freezing resistance. Even so, their collagen networks will irreversibly aggregate alongside growing ice crystals, ultimately leading to the tough and rubbery texture commonly seen in poorly frozen cephalopod products (Schmidt and Mouritsen 2022). Thus, the key regulating factor here lies not in ice crystal size, but in the rate of collagen denaturation. In practical application, solutions aimed at solving one specific quality problem cannot be directly applied to other aquatic products. Magnetic field treatment that slows down blackening in shrimp fails to inhibit formaldehyde accumulation in cod. Meanwhile, ultrasound parameters optimized to stabilize tuna lipids may even worsen texture defects in collagen‐rich octopus. This species‐specific difference will be fully discussed in the following sections of this review.

3. Laws of Ice Crystal Formation and Crystal Growth During Freezing and Their Correlation With Aquatic Product Quality

3.1. Ice Crystal Nucleation and Growth

Ice crystal formation consists of two core phases: nucleation and subsequent growth. Both processes are governed by thermodynamic and kinetic factors, whose combined effects decide the final shape, size, and distribution of ice crystals inside aquatic tissues. In thermodynamic terms, water freezing is a phase transition process driven by Gibbs free energy change. When the temperature drops below 0°C, forming regular hydrogen‐bonded ice structures cut down enthalpy more than the energy consumed by reduced molecular activity. This makes ice more stable than liquid water and offers the basic driving force for natural water freezing (Parandi et al. 2022; Roos 2021). Yet liquid water turning into ice needs to break through the nucleation energy barrier, which comes from surface tension between newly formed ice nuclei and surrounding water. Stable ice nuclei can only form when energy released during ice generation surpasses interfacial energy. Pure water needs severe supercooling ranging from −30°C to −40°C to achieve homogeneous nucleation, and this phenomenon hardly occurs in actual food freezing scenarios (Tan et al. 2021a).

Within food systems, specifically in the case of aquatic tissues, ice crystals predominantly undergo heterogeneous nucleation at internal interfaces, including intermyofibrillar spaces, sarcolemmal phospholipid bilayers, and dispersed protein/mineral particles. This phenomenon elevates the apparent nucleation temperature to the range of −1°C to −4°C (Pérez‐Bermúdez et al. 2023). Despite a substantial amount of empirical research, the molecular identity of the dominant nucleation sites in muscle tissue remains indeterminate. The classical “lattic‐match” theory fails to elucidate why ice nucleates on specific biopolymer surfaces while not on other materials with comparable crystallographic registry. Emerging evidence suggests that local interfacial water structuring is a more decisive factor (Xie et al. 2023). This unresolved issue continues to constrain the design of nucleation‐control strategies in muscle foods.

From a kinetic perspective, the ice–crystal growth rate is jointly influenced by temperature gradient, supercooling, and water diffusivity, with the moving solid–liquid front conventionally modeled as a Stefan problem—a classical formulation that posits a positive correlation between growth rate, local supercooling (the driving force for molecular migration), and thermal diffusivity (which controls latent‐heat removal at the interface) (Parandi et al. 2022). However, this idealization breaks down in heterogeneous muscle tissue: an anisotropic heat conduction along fiber bundles, solute partitioning at the advancing front, and the visco‐elastic resistance of the protein matrix, all introduce systematic deviations from classical predictions, and a unified kinetic model that fully incorporates these biological complexities is still lacking (Huo et al. 2024).

Building on this thermodynamic and kinetic framework, freezing regimes in muscle tissue are conventionally partitioned into three operational categories that differ in the balance between supercooling and latent‐heat removal at the advancing front. Under slow freezing (<0.5 cm/h), ice crystals grow at a slow pace. Solutes (salts, amino acids, and low‐molecular carbohydrates) are continuously excluded from the lattice as they cannot be incorporated into the ordered hexagonal structure of ice I (hexagonal ice, the stable ice phase under atmospheric pressure). This exclusion gradually concentrates the unfrozen liquid phase, thereby lowering its freezing point and reducing the local supercooling between adjacent crystals. Ultimately, only a few nuclei in favorable positions expand into large dendritic or needle‐like crystals (Tan et al. 2021a). The cryo‐concentrated unfrozen phase resulting from this segregation also governs subsequent frozen‐storage stability through its glass‐transition temperature T′_g, a state‐diagram‐based control parameter that is seldom emphasized in conventional descriptions of slow freezing but is increasingly recognized as crucial for long‐term protein stability (Li, Wang, Yanagita, et al. 2024; Roos 2021).

At intermediate freezing rates (0.5–2 cm/h), the heat and mass transfer at the moving front are more closely balanced, and the latent‐heat release partially alleviates local supercooling. The nucleation density increases compared to slow freezing, whereas anisotropic growth along the temperature gradient is maintained, resulting in columnar crystals whose long axis is oriented parallel to the heat‐flow direction. This intermediate morphology is mechanistically different from both the dendritic crystals of slow freezing and the equiaxed crystals of rapid freezing, and its parameter range defines the operating window that most physical‐field interventions discussed in Section 4 aim to shift towards finer, more uniform structures (Pérez‐Bermúdez et al. 2023; Yan et al. 2023).

In contrast, rapid freezing (>2 cm/h) causes the entire system to reach a high degree of supercooling before significant solute segregation can occur, triggering the simultaneous nucleation of numerous small ice crystals (10–50 µm). The mutual spatial confinement between adjacent nuclei suppresses dendritic extension and results in a large number of fine, uniform equiaxed crystals (Yan et al. 2023; Yang, Ye, et al. 2023).

However, the prevailing assumption that “smaller ice crystals always yield better quality” is not universally valid. Liquid‐nitrogen immersion at −125°C to −196°C produces crystals below 20 µm but frequently induces macroscopic surface cracking in fish fillets and shrimp. This results from the steep thermal‐stress gradient that exists between surface and core, leading to recommendations that the effective freezing temperature for industrial fish freezing be capped near −80°C even when further smaller crystals could be achieved at much lower temperatures (Xie et al. 2023; Yan et al. 2023). The optimum freezing rate therefore reflects a trade‐off between intracellular ice damage and bulk thermal‐stress fracture—a trade‐off seldom quantified in mechanistic models (Zhang et al. 2024).

3.2. Recrystallization

Recrystallization during frozen storage is a composite phenomenon involving three concurrent mechanisms rather than the single Ostwald‐ripening process commonly invoked: (i) migratory recrystallization (Ostwald ripening), driven by the Gibbs–Thomson chemical‐potential gradient, in which water diffuses from small high‐curvature crystals to larger crystals; (ii) accumulative recrystallization, in which adjacent crystals coalesce; and (iii) isomass recrystallization, in which crystal surfaces smooth without net mass exchange to minimize surface energy (Shang et al. 2024; Tan et al. 2022). Their relative contributions depend on the frozen storage temperature and temperature fluctuation, water content, and the initial ice–crystal distribution.

Under realistic frozen‐storage conditions, periodic temperature fluctuations (±2–3°C) intensify all three mechanisms: Surface melting during the warm half‐cycle releases water that refreezes onto larger crystals on cooling, generating a “melt–refreeze” loop that progressively coarsens the ice crystal population (Liu, Zeng, et al. 2025; Tan et al. 2022). At −18°C, the average ice–crystal size in frozen fish has been reported to increase by 30%–40% within the first 30 days and by 60%–80% over 90 days; however, the magnitude is strongly species‐, formulation‐, and packaging‐dependent, and the figures cannot be generalized across products (Yang, Ye, et al. 2023).

A control parameter that is mechanistically central yet rarely emphasized is the glass‐transition temperature of the maximally freeze‐concentrated unfrozen matrix, T′_g, which for fish muscle typically lies between −10°C and −13°C. Below T′_g, the freeze‐concentrated phase enters a kinetically arrested state in which long‐range translational diffusion in the amorphous matrix is slowed by several orders of magnitude, suppressing—but not abolishing—recrystallization and protein‐aggregation processes (Lee, Jo, et al. 2024; Roos 2021). Most commercial cold chains operating at −18°C, therefore, sit only 5–8°C below T′_g, a margin that is insufficient for true kinetic arrest for three interlocking reasons: (i) sub‐T′_g secondary (β‐) relaxations of the unfrozen matrix and self‐diffusion at ice–crystal surfaces continue at measurable rates and provide a residual pathway for crystal coarsening; (ii) real frozen muscle rarely attains maximal freeze‐concentration, so a fraction of the unfrozen water remains in a rubbery state with an effective local T′_g below the nominal −10°C to −13°C value; and (iii) the routine ±2–3°C cold‐chain temperature fluctuations periodically lift the matrix above T′_g, where translational mobility recovers exponentially and is not regained on cooling. Slow recrystallization and protein‐aggregation drift are therefore kinetically inevitable rather than merely possible under standard −18°C storage. Only ultra‐low‐temperature storage at ≤−25°C to −30°C—that is, 12−20°C below T′_g rather than 5–8°C—approaches genuine molecular immobilization (Lee, Jo, et al. 2024; Roos 2021).

The impact of recrystallization on the overall quality of aquatic products manifests in three coupled dimensions:

1. Aggravated thawing loss as larger crystals raise the drip rate from 5%–10% to 20%–30%;

2. Depletion of water‐soluble nutrients (proteins, free amino acids, and vitamins);

3. Progressive disruption of sarcolemmal and myofibrillar architecture, with the increased microscopic porosity simultaneously enhancing oxygen ingress and accelerating lipid and protein oxidation (Fan et al. 2024; Xie et al. 2023).

Strategies to suppress recrystallization include constant ultra‐low‐temperature storage (preferably below T′_g, i.e., ≤−25°C), tight thermal control (within ±1°C), physical‐field‐assisted nucleation, and the incorporation of cryoprotectants such as sugars, polyols, polysaccharides, and antifreeze peptides (Fan et al. 2024; Li et al. 2023). However, comparative trials reveal substantial heterogeneity in cryoprotectant performance. The performance efficacy can vary by an order of magnitude across fish species, peptide sequence, and freeze–thaw protocol, and no agent has been shown to be universally effective. Whether this heterogeneity reflects genuine matrix‐specific physicochemistry or methodological inconsistencies between studies remains an open question that limits translational deployment (Xu et al. 2024).

3.3. Correlation Between Ice Crystal Properties and Aquatic Product Quality

Ice–crystal morphology is typically classified into dendritic, needle‐like, columnar, and equiaxed types based on the freezing rate. Although this discrete taxonomy is convenient for educational purposes, it oversimplifies the situation of frozen muscle, where transitional and mixed morphologies coexist within a single sample due to the inherent spatial gradients in cooling rate between the surface and the core (Pérez‐Bermúdez et al. 2023). Dendritic crystals (branched, tree‐like) form at very low freezing rates (<0.5 cm/h); needle‐like crystals emerge under moderately slow freezing; columnar crystals are generated at intermediate rates (0.5–2 cm/h) where heat and mass transfer are in balance; and equiaxed crystals form at sufficiently high rates (>2 cm/h) to inhibit directional growth.

During the slow freezing of aquatic products (<0.5 cm/h), ice crystals are usually dendritic or needle‐like (100–300 µm), growing along the thermal gradient through intermyofibrillar spaces and even extending into the sarcoplasm. Their sharp edges and branches pierce the sarcolemma (leading to membrane rupture and the leakage of intracellular K⁺ and myoglobin), compress and shear myofibrils, and dissociate sarcomeric Z‐lines, resulting in severe mechanical damage (Tan et al. 2021a). In contrast, equiaxed crystals (10–50 µm) generated by rapid freezing are confined to interstitial spaces and intermyofibrillar gaps; spatial constraints from neighboring nuclei prevent excessive expansion, so the myofibrillar architecture and sarcolemmal integrity remain largely intact, and water reabsorption upon thawing is favored (Huo et al. 2024; Yan et al. 2023).

Quantitative correlations between ice–crystal size and quality indices, including the often‐cited “thawing loss increases by 3–5 % per 50 µm increase in mean crystal size,” originate from a limited number of species‐specific studies. Regression slopes vary considerably between fish species, fillet thickness, freezing medium, and storage history; pooled meta‐analysis is currently lacking, and the figure should not be treated as a universal constant (Yang, Ye, et al. 2023). Texture and water‐holding capacity (WHC) additionally depend on collagen‐network integrity and MP conformation, both of which can be partly decoupled from ice–crystal geometry per se (Shang et al. 2024). In addition, nonuniform ice–crystal distribution with small crystals near the surface and large crystals at the center, which arises from the unavoidable through‐thickness temperature gradient, produces a “firm‐outside, loose‐inside” texture after thawing that significantly degrades sensory quality (Liu et al. 2025).

Quantitative characterization of ice–crystal morphology relies on several complementary techniques: light microscopy after freeze‐substitution or freeze‐drying, cryo‐scanning electron microscopy (SEM), x‐ray micro‐CT, DSC for total ice content, and NMR/MRI for nondestructive in situ tracking, each carrying distinct methodological constraints. Freeze‐substitution and freeze‐drying, the most common preparation routes, are themselves prone to artifacts: Solvent exchange can alter the apparent porosity, and re‐warming during sectioning can promote secondary recrystallization. So the reported “ice‐crystal sizes” should be interpreted as method‐conditional rather than absolute values (Pérez‐Bermúdez et al. 2023). Cross‐laboratory comparisons that explicitly account for these preparation effects remain rare, which partially explains the inconsistency of quantitative correlations reported in the literature.

The practical implications for physical‐field freezing should therefore be product‐specific rather than generic. For fresh‐retail products such as whole shrimp and fillets, refining crystal size to 10–30 µm using ultrasound‐ or high‐pressure‐assisted freezing can lower thawing drip below 10% (Fan et al. 2024; Ma et al. 2022); however, the dose–response curve is non‐monotonic, as excessive ultrasonic power (>250 W) has been shown to enlarge rather than refine crystals owing to localized heating and cavitation‐induced mechanical stress, making empirical optimization unavoidable (Ma et al. 2022). For gel‐type processed products, such as surimi, fish balls, and crab sticks, preserving MP gelation capability is more critical than minimizing crystal size. HPAF has been shown to simultaneously refine ice crystals and partially densify the gel network, reducing thawing loss by ∼50% and protein carbonyl formation by ∼35% across multiple freeze–thaw cycles (Li et al. 2023; Liu et al. 2025). Converting these product‐specific results into a generalizable design rule remains a major outstanding problem. This evidence indicates that ice–crystal size should be treated as a necessary but insufficient quality indicator. Small and uniform crystals usually reduce mechanical disruption, but they do not automatically prevent protein unfolding, lipid oxidation, thermal cracking, or storage‐induced recrystallization. Therefore, studies that report only crystal refinement without parallel evidence on WHC, protein stability, oxidation, and sensory quality provide an incomplete assessment of freezing performance. This is a key reason why apparently positive results from different freezing technologies cannot be directly compared or generalized across aquatic products.

4. Physical‐Field‐Assisted Freezing Technologies and Related Approaches

The thermodynamic and kinetic limits established in Section 3 define a tripartite engineering target: simultaneously (i) suppress the heterogeneous‐nucleation barrier so that nucleation becomes spatially homogeneous, (ii) terminate dendritic post‐nucleation growth at the equiaxed/columnar boundary, and (iii) suppress translational mobility in the cryo‐concentrated unfrozen matrix without perturbing the molecular machinery underlying texture and flavor—a function that requires either storage well below T′_g (≤−25°C to −30°C, i.e., 12–20°C sub‐T′_g rather than the 5–8°C margin offered by −18°C cold chains) or in‐process interventions that constrain residual sub‐T′_g diffusion (e.g., ice‐surface refinement, cryoprotective immobilization of the unfrozen film). No truly heat‐transfer‐based process can satisfy these constraints at once, because each is governed by an independent physical lever with the chemical potential of supercooled water, hydrodynamics at the freezing front, and translational mobility in the unfrozen amorphous phase. Physical‐field‐assisted freezing reframes the problem by introducing a nonthermal control variable that intervenes directly at one of these levers—pressure, acoustic cavitation, magnetic flux, electric polarization, or dielectric heating. The taxonomy that follows departs from the conventional “list‐by‐technology” structure of earlier reviews; instead, each archetype is interrogated against four binding criteria, namely, dominant physical mechanism and the strength of mechanistic evidence in aquatic muscle; decisive process variables with their boundary windows; cross‐species robustness; and engineering envelope. Section 4.7 then synthesizes these readings into a unified parameter–outcome map (Figure 2) and benchmarks the integrated quantitative evidence reported in Tables 1 and 2.

FIGURE 2.

Physical‐field‐assisted freezing mechanism and quality regulation in aquatic products. EFAF, electric field‐assisted freezing; HPAF, high‐pressure‐assisted freezing; MFAF, magnetic field‐assisted freezing; MWAF, microwave‐assisted freezing; PSF, pressure‐shift freezing; RFAF, radiofrequency‐assisted freezing; UAF, ultrasound‐assisted freezing; US‐HPAF, ultrasound‐assisted high‐pressure freezing.

TABLE 1.

Comparative parameter–outcome matrix for high‐pressure‐assisted freezing (HPAF)/pressure‐shift freezing (PSF), high‐pressure thawing (HPT), and ultrasound‐assisted freezing (UAF) in aquatic products under standardized reporting (frequency in kHz, power in W or W/cm², pressure in MPa, time in min).

Technology	Species (binomial)	Decisive parameters	Freezing medium/Rate	Ice crystal size (µm)	Drip/Thawing loss	Protein/Lipid index	Industrial feasibility	References
PSF	Salmo salar (fillet 50 mm)	200 MPa, −20°C, release ≥25 MPa/s	Indirect, ∼1.2 cm/h	15–35 (vs. >150)	14% → 6%	Myosin stable ≤200 MPa	Premium fillets, batch 50–100 L	Li et al. (2022)
HPAF	Trachinotus ovatus fish balls	200 MPa, −20°C	Indirect	Refined; slowed recryst	0.68× control over 5 F/T	Carbonyl ↓34.5%, TBARS ↓15.6%	Surimi‐type products	Liu et al. (2025)
HPAF	Sepia officinalis mantle	150–200 MPa, −20°C	Indirect	30–60	15% → 9%	Collagen toughening at >250 MPa	Cephalopod fillets	Nilsuwan et al. (2024)
HPT	Pampus argenteus	100–150 MPa thawing	—	n/a	9% → 5% (water loss ↓6.8%)	Color ΔE ↓56.8%	Industrial thawing	Cui et al. (2019)
MUIF	Larimichthys crocea (fillet 25 mm)	30 kHz tri‐frequency, 175 W, 5 min, 5/5 s pulse	Brine immersion, 2.4 cm/h	18–35	Drip ↓46% over 90 days	Ca2+‐ATPase preserved	IQF tunnels	Ma et al. (2021)
MUIF (long‐storage)	L. crocea	175 vs. 190 W	Brine	—	—	TBARS ↑1.8× at 190 W	Power‐window critical	Yang et al. (2024)
UIF	Cyprinus carpio (fillet)	28 kHz, 175 W, 6 min	Brine, 2.1 cm/h	22–45	Cooking loss ↓28%	T 21/T 22 shortened	Freshwater fillets	Sun et al. (2019)
UAF	Crassostrea gigas	20 kHz, 0.2 W/cm2, 10 min	Brine	25–50	18% → 10%	Inverted dose–response	Shellfish IQF	Suárez‐Medina et al. (2024)
UAF (orthogonal)	Argyrosomus japonicus	Dual‐freq 28 + 40 kHz	Brine	Smallest at intermediate W	Drip ↓25%	Protein structure preserved	Fillet IQF	Yu and Xie (2023)
LNF	Litopenaeus vannamei	−95°C immersion	LN2 spray	8–18	Lowest at −95°C; cracks at −125°C	WHC max	Premium IQF	Yan et al. (2023)
LNF	Portunus trituberculatus	−100°C, 4‐month storage	LN2	Refined	TBARS ↓ for 4 months	Optimal warmer than fish	Crustacean shelf‐life	Ren et al. (2025)
Cryo + ABF	Oreochromis niloticus	LN2 + ABF (combined)	Mixed	Refined surface, larger core	5.32 vs. 268 min freeze time	90‐day storage	Industrial hybrid	Regalado et al. (2024)

Abbreviations: ABF, air‐blast freezing; LNF, liquid‐nitrogen freezing; MUIF, multifrequency ultrasound‐assisted immersion freezing; UIF, ultrasound‐assisted immersion freezing.

TABLE 2.

Standardized minimum reporting matrix and reported parameter ranges for magnetic field‐assisted freezing (MFAF), electric field‐assisted freezing (EFAF), high‐voltage electrostatic field (HVEF), radiofrequency‐assisted freezing (RFAF), microwave‐assisted freezing (MWAF), and hybrid configurations in aquatic products.

Technology	Decisive parameters	Reported range (aquatic)	Crystal/Quality effect	Evidence levela	Dominant limitation	References
SMF	Flux density (mT); time (min); and uniformity	0.5–3 mT (5–30 G); 30–180 min	Non‐monotonic; peak at 15 G in tilapia	Low	Reproducibility; <100 mT thermodynamically marginal	Wei et al. (2021)
AMF	Flux (mT); frequency (Hz); waveform	5 mT; 50–250 Hz, peak 200 Hz	Crystal ↓; FD ↑; α‐helix preserved	Low–Mod	Coupling geometry, waveform decisive	Ye et al. (2024)
SMF/AMF	Waveform vs. intensity in catfish fillet	5 mT; static + alt	Phase‐transition time ↓; protein altered	Low–Mod	Waveform > intensity	Yang et al. (2024)
AMF	Flux (mT) in shrimp	5–10 mT	Architecture preserved	Low	Cuticle artifact unresolved	Yang, Zhang, et al. (2025)
Superconducting MF	Flux (T)	6–10 T (saline/gel models)	Reproducible nucleation T shift	Mod	Not industrially feasible	Wei et al. (2021)
MF mechanism review	Spin/H‐bond/Lorenz	Multi‐source	Theoretical models incomplete	Mod	Mechanism contested	Pu et al. (2025)
PEF pretreat.	Field (kV/cm); pulse µs; n pulses	0.6–1.0 kV/cm; salmon	Total loss ↓6%; lipid oxidation ↑slight	Mod	Membrane electroporation >1.5 kV/cm	Li et al. (2020)
LF‐EFAF	Field (kV/m)	15–20 kV/m; tilapia/shrimp	Freezing rate ↑; crystal refined	Low–Mod	Field attenuation in saline tissue	Sun et al. (2024)
OMF + PEF supercool	OMF 15 mT + PEF 8 V	salmon −7°C supercooled	TBARS = refrig.; microbial low	Mod	Stability of supercooled state	Lee et al. (2024)
HVEF (thawing)	Voltage (kV); gap (cm)	4.5–14 kV; 3–6 cm; tilapia	Thaw 2.16× faster	Mod	Protein solubility ↓ at high V	Hafezparast‐Moadab et al. (2018)
HVEF + preserv.	2.5–4.0 kV combined	rainbow trout	Shelf‐life ↑ vs. single	Mod	Salt distribution heterogeneity	Qin et al. (2025)
RFAF	Frequency (MHz); power (W/kg); pulse	13.56–27.12 MHz; trout	Drip ↓; sensitive to gap/pulse	Low–Mod	Penetration <40 mm; ice fraction reduces	Hafezparast‐Moadab et al. (2018)
MWAF	Frequency; power; and duty	915/2450 MHz; pork → fish proxy	Crystal ↓62% (pork)	Low	Standing waves; no aquatic confirmation	Anese et al. (2012)
US‐HPAF (hybrid)	MPa + kHz + W/cm	200 MPa + 25 kHz + 0.2 W/cm2	Drip 7% → 4% in shrimp	Mod	Reactor complexity	Hu et al. (2022)
PEF–ABF (hybrid)	PEF + ABF	0.6 kV/cm + ABF; salmon	PSF‐comparable refinement, 1/5 energy	Mod	Geometry‐sensitive	Rondineli and Silva (2024)

Abbreviations: ABF, air‐blast freezing; AMF, alternating magnetic field; SMF, static magnetic field; US‐HPAF, ultrasound‐assisted high‐pressure freezing.

^a Evidence: Low = single‐laboratory or unreplicated; moderate = ≥2 independent laboratories with consistent direction; high = multi‐laboratory consensus + mechanistic confirmation. None of the technologies in this table currently reaches “High” in aquatic‐muscle systems.

4.1. HPAF and PSF

HPAF and PSF exploit the depression of the ice I liquidus approximately 1°C per 10 MPa, reaching −21°C at ∼210 MPa (the ice I/III/liquid triple point near 209 MPa, beyond which higher density ice polymorphs (II, III, and V) become accessible (Otero and Sanz 2003). The two operational modes are mechanistically distinct: HPAF nucleates and grows ice under sustained pressure, whereas PSF cools the sample to a deeply metastable supercooled liquid at 100–200 MPa and then triggers system‐wide nucleation by the rapid depressurization, converting the released supercooling into a near‐uniform burst of fine equiaxed crystals (Li et al. 2022).

Three parameters dominate the outcomes as follows: pressure level, depressurization rate, and the depth of supercooling at release. Across aquatic studies, the optimal PSF window converges on 150–200 MPa with rapid release (typically tens of MPa/s). In Salmo salar (Atlantic salmon) and Micropterus salmoides (largemouth bass), PSF generated intracellular ice clusters of 15–35 µm versus >150 µm under air‐blast freezing (ABF), halving the thawing drip and preserving WHC over 60–90‐day frozen storage at −18°C (Jia, Roy, et al. 2022; Li et al. 2022). For Trachinotus ovatus (golden pomfret) fish balls, HPAF at 200 MPa reduced thawing loss fraction to 0.68 compared with the heat‐treated control and lowered carbonyl increase by 34.5% and thiobarbituric acid reactive substances (TBARS) by 15.6% over five freeze–thaw cycles, whereas microstructural analysis confirmed altered crystal geometry and slowed recrystallization (Liu et al. 2025). The dose–response, however, is non‐monotonic above 250–300 MPa: Hydrostatic compression denatures myosin and actin, exposes hydrophobic surfaces, and reduces Ca²⁺‐ATPase activity by 30%–60% (Roobab et al. 2022). Myosin destabilization occurs from 200 MPa and actin from 300 MPa in S. salar (Russo et al. 2024), placing the engineering window squarely between crystal‐refinement and protein‐denaturation regimes.

Two mechanistic ambiguities remain in aquatic systems. First, it remains unresolved whether the benefits of PSF primarily stem from homogeneous nucleation in the depressurized liquid or from the polymorphic conversion of metastable ice III/V to ice I. Molecular‐dynamics studies support concurrent contributions, yet in situ structural evidence in muscle is lacking (Roos 2020). Second, the partial cryo‐resistance conferred to cephalopod collagen by moderate pressure is offset at higher pressure by aggregation around ice fronts. An interaction has been documented for Loligo spp. and Dosidicus gigas (Humboldt squid), where treatment at 200–400 MPa modified the cuticular and connective‐tissue architecture (Nilsuwan et al. 2024; Zura‐Bravo et al. 2025). The technological limitation persists as industrial throughput: vessels with a capacity above 100 L rated for 600 MPa cycling are commercially available but necessitate a capital expenditure of $1.5–$3 million and operate in batch mode (Pou 2021). High‐pressure thawing (HPT), the inverse process, melts the product from the core outwards; for Pampus argenteus (silver pomfret), HPT at 100–150 MPa was reported to reduce color deviation by 56.8% and water loss by 6.8% compared to conventional thawing (Cui et al. 2019).

4.2. Ultrasound‐Assisted Freezing

UAF integrates conventional immersion or contact freezing with low‐frequency, high‐intensity ultrasound (typically 20–40 kHz, 0.1–0.5 W/cm²) delivered either continuously or in a pulsed manner. Acoustic cavitation facilitates heterogeneous nucleation via the transient collapse of micro‐bubbles. The subsequent rebound of these micro‐bubbles generates local pressure pulses, which elevate the equilibrium freezing temperature and induce secondary nucleation by fragmenting nascent dendrites (Cheng et al. 2017; Yu and Xie 2023). Microstreaming disrupts the boundary layer at the freezing front, increasing the effective heat‐transfer coefficient by a factor of 2–5 and yielding equiaxed crystals with a size range of 10–40 µm.

The crucial parameters in this context are frequency, power density, treatment duration, and duty cycle. Multifrequency ultrasound‐assisted immersion freezing (MUIF) has emerged as a significant improvement over single‐frequency UAF. In Larimichthys crocea (large yellow croaker), tri‐frequency MUIF achieved the highest cavitation efficiency and the shortest freezing time, resulting in the smallest and most uniform crystals, as well as the best preservation of myofibrillar structure (Bian et al. 2022; Ma et al. 2021, 2022). At 175 W applied power, MUIF reduced thawing loss by 46% and protected lipid stability beyond 90 days at −18°C, whereas at 190 W, cavitation‐induced radicals raised TBARS to levels exceeding the unsonicated control (Yang, Dong, et al. 2024). For Cyprinus carpio (common carp) muscle, ultrasound‐assisted immersion freezing at 175 W minimized pore diameter, shortened T ₂₁/T ₂₂ relaxation, and preserved emulsifying capacity of MP during frozen storage (Sun et al. 2019). For Macrobrachium rosenbergii (giant river prawn), MUIF mitigated mechanical damage and protein deterioration at intermediate powers (Ren et al. 2025), and for Sciaenops ocellatus (red drum), UAF improved muscle quality through controlled MP denaturation (Qiu et al. 2022). Multifrequency operation broadens the cavitation spectrum and reduces the formation of acoustic dead zones in the immersion bath, which is thought to underlie the improved uniformity over single‐frequency UAF (Bian et al. 2022; Ma et al. 2022).

The non‐monotonic dose–response is a defining feature of UAF. In Argyrosomus japonicus (sea bass), orthogonal dual‐frequency at intermediate amplitude resulted in the formation of the smallest ice crystals. However, excessive power reversed the trend via localized heating, which led to the partial remelting of nascent crystals (Yu and Xie 2023). Independent research on Crassostrea gigas (Pacific oyster) and Argopecten irradians (bay scallop) indicated that matrices with high lipid or protein content invert the conventional dose–response relationship, shifting the optimum towards lower power and higher frequency (Li, Wang, Zeng, et al. 2024). Hydroxyl and superoxide radicals derived from cavitation accelerate the oxidation of phospholipids. This effect was quantified in L. crocea treated with MUIF, where the TBARS increased by 1.8‐fold at 0.5 W/cm² over a 30‐day period at −18°C (Yang, Dong, et al. 2024). Acoustic attenuation in muscle (1–3 dB/cm at 25 kHz) constrained penetration to <30 mm, limiting UAF to fillets and individually quick frozen (IQF) products rather than thick blocks (Zhang et al. 2018). Operating energy is favorable (0.05–0.10 kWh/kg above baseline), and continuous‐belt UAF tunnels rated for ≥500 kg/h are commercially deployed for shellfish (Liu, Liao, et al. 2022).

4.3. Magnetic Field‐Assisted Freezing

Among the field‐assisted technologies, MFAF has attracted the largest commercial interest while remaining the most poorly understood mechanistically. The proposed mechanism, namely, proton‐spin polarization, hydrogen‐bond network reordering, and reduced Lorenz forces between water clusters, has been embedded in commercial Cells Alive System (CAS) freezers and similar units but rests on weak first‐principles support (Pu et al. 2025). Theoretical estimates place the magnetic contribution to ΔG at <10⁻⁴ kJ/mol at 1 T, several orders of magnitude below any thermodynamically meaningful nucleation barrier (Otero et al. 2018).

Empirical evidence in aquatic products is heterogeneous but instructive when regrouped by field topology. Static magnetic fields (SMFs) at 0.5–3 mT applied to Oreochromis niloticus (Nile tilapia) produced freezing‐time shortening and crystal‐size reduction with a non‐monotonic intensity dependence peaking at 1.5 mT (Wei et al. 2021). Alternating magnetic fields (AMF) at 5 mT, 200 Hz minimized crystal size and maximized fractal dimension (FD) in Oreochromis fillets while preserving α‐helix and β‐sheet content (Ye et al. 2024). For Ictalurus punctatus (channel catfish), SMF and AMF shortened phase‐transition time and reduced crystal size, with waveform, not intensity, emerging as the decisive variable (Yang, Yang, et al. 2024). For Marsupenaeus japonicus (kuruma shrimp), AMF preserved muscle architecture and water status more effectively than ABF (Yang, Zhang, et al. 2025). Confirmation has also come from nonaquatic muscle (Gan et al. 2024; Liu, Heming, et al. 2025), supporting that the effect, while small, is not zero.

Two diagnostics suggest that observed effects, where reproducible, are most likely indirect, mostly driven by residual mechanical vibration of solenoids, eddy‐current heating of coils, or convective effects in the freezer cavity, rather than by direct quantum‐mechanical action on water. First, the magnitude of reported supercooling extension (1–3°C) lies within the range achievable by passive supercooling in well‐controlled freezers without any field (Lee, Tang, et al. 2024). Most experimental food‐freezing studies have been confined to permanent‐magnet or solenoid systems below 200 mT, and recent comprehensive reviews highlight that reproducibility across these low‐field studies remains poor and the underlying mechanisms are still unclear (Yang, Tian, et al. 2025; Pu et al. 2025). Until these confounders are systematically excluded in future work, the magnitude of any direct magnetic effect remains uncertain.

4.4. EFAF and Pulsed Electric Field Pretreatment

EFAF encompasses static (SEF, 0.1–10 kV/cm), pulsed (PEF, µs pulses), high‐voltage electrostatic (HVEF, 2.5–14 kV with corona discharge), and oscillating regimes. The dominant proposed mechanism is dipole alignment of water molecules and lowering of the nucleation barrier through field‐induced reorientation that disturbs hydrogen‐bond topology. This is supported mainly by molecular‐dynamics simulations and remains a subject of active mechanistic debate (Wang et al. 2025; Peng et al. 2025). Recent simulation work shows that local interfacial electric fields, rather than lattice match, regulate heterogeneous nucleation, providing first‐principles support for the EFAF concept but not evidence of effect at industrially accessible field strengths (Yu et al. 2025). The interaction between field and the high ionic strength of muscle (∼0.15 M) attenuates the field within the sample interior, complicating interpretation of “applied” versus “effective” field strength (Peng et al. 2025).

Empirical aquatic data show modest, parameter‐sensitive benefits. PEF pretreatment of S. salar fillets at 1.0 kV/cm before freezing was reported to reduce thawing time by 20 min (−2°C to 0°C) and total loss by 6%, but the b* value and lipid peroxidation were elevated through PEF‐induced oxidation (Li et al. 2020). Low‐frequency EFAF of Oreochromis fish protein produced finer ice and improved structural integrity (Sun et al. 2024). For deep‐water shrimp (Solenocera melantho), HVEF‐assisted freezing at 15–20 kV/m accelerated passage through the maximum ice‐formation zone and produced smaller, more uniform ice crystals, with concomitant improvements in muscle texture and reductions in MP degradation (Liu, Wang, et al. 2022). Combined OMF + PEF supercooling held S. salar at −7°C without freezing while keeping TBARS at refrigerated levels and microbial counts below frozen controls (Lee, Tang, et al. 2024).

HVEF excels primarily for thawing rather than freezing. Regarding Oreochromis fillets, HVEF accelerated the thawing process by a factor of 2.16 compared to the conventional control. Nevertheless, an increase in voltage led to a reduction in protein solubility and an acceleration of oxidation (Hafezparast‐Moadab et al. 2018). In the case of T. ovatus, HVEF salting at 3.5 kV optimized the salt distribution while constraining protein and lipid oxidation (Chen, Wang, Gao, et al. 2025). For Oncorhynchus mykiss (rainbow trout), the combination of HVEF and composite preservatives extended the ice‐temperature shelf‐life more effectively than either intervention used independently (Qin et al. 2025). In active freezing of aquatic products, electric‐field‐assisted freezing and PEF‐based pretreatments currently merit only a moderate evidence rating. Most reported benefits remain parameter‐sensitive, and the evidence base for direct electric‐field control of ice nucleation in aquatic muscle is still limited (Çalışkan Koç et al. 2025; Peng et al. 2025).

4.5. RFAF and MWAF

RFAF (13.56 or 27.12 MHz) and MWAF (915 or 2450 MHz) exploit the low‐power dielectric heating of the unfrozen aqueous fraction to maintain a dynamic, controlled supercooling regime. Brief, low‐power pulses partially remelt nascent dendrites at the freezing front, suppressing dendritic growth and favoring equiaxed crystals (Anese et al. 2012; Sadot et al. 2017). The mechanism is conceptually attractive: The dielectric loss factor of muscle drops sharply as ice fraction rises, so heating selectively targets the unfrozen interfacial film around growing crystals—a regulation pathway physically distinct from pressure or acoustic intervention.

Aquatic‐product evidence remains skewed toward thawing rather than active freezing. For O. mykiss, RFAF reduced drip loss versus air‐blast and cryogenic flow, but treatment efficacy depended sensitively on RF pulse pattern and electrode gap (Hafezparast‐Moadab et al. 2018). For Penaeus vannamei and Euphausia superba, dielectric properties were systematically mapped, providing the basis for designed RF/MW protocols (Yang et al. 2017). MWAF of pork tenderloin has been reported to reduce the average ice–crystal size by approximately 62% relative to conventional freezing. However, the direct transfer of these results to aquatic muscle is uncertain because the dielectric loss of fish flesh below −5°C differs from that of pork. Additionally, the fibrous architecture of fish is more anisotropic, and MWAF of muscle remains at the proof‐of‐concept stage (Wang et al. 2025). For Thunnus albacares (yellowfin tuna) blocks, RF thawing at 27.12 MHz halved thawing time relative to water immersion (Yang et al. 2019). The principal limitations are nonuniform field distribution (standing waves, edge heating), the rapid drop of dielectric loss as the ice fraction grows, and the difficulty of integrating RF/microwave applicators into continuous freezing tunnels without compromising thermal isolation (Wang et al. 2025). Studies on Atlantic salmon (S. salar) reveal that deep freezing aided by ferroferric oxide magnetic nanoparticles paired with microwave thawing can well maintain the integrity of muscle fiber tissues. This treatment keeps the muscle fragmentation index as low as 2.73 ± 0.31, stabilizes the secondary and tertiary structures of proteins, and strengthens hydrogen bonding between protein molecules and water. These changes suggest that this method can effectively alleviate tissue damage caused by repeated freezing and thawing (Wang, Bu, et al. 2024).

4.6. Cryogenic and Hybrid Field Approaches

Although LNF is conventionally classed as a heat‐transfer‐only method, it interacts with the physical‐field portfolio as a benchmark and as a hybrid building block. In Litopenaeus vannamei, LNF at −95°C minimized pore size and drip loss, whereas −125°C induced cryogenic surface fracture (Yan et al. 2023). For Portunus trituberculatus (gazami crab), LNF at −100°C optimally inhibited oxidative deterioration and extended frozen shelf‐life to 4 months—notably warmer than the optimum for shell‐free fish, reflecting crustacean‐specific cuticle fracture risk (Ren et al. 2025). In O. niloticus, combined cryogenic + ABF achieved 5.3‐min freezing time versus 268 min for static air, with downstream quality maintained over 90 days at −18°C (Regalado et al. 2024). In Perca fluviatilis, label‐free proteomics distinguished LNF, IF, and AF through differentially abundant protein markers (Zhang et al. 2024). A possible link to ferroptosis‐related lipid peroxidation has been proposed (Huang et al. 2024; Zheng et al. 2026), although direct experimental support in aquatic muscle is currently limited.

Hybrid configurations are currently the most active area of research. The underlying idea is that the techniques act on different stages of the freezing process: pressure on nucleation thermodynamics, ultrasound on the multiplication and dispersion of nuclei, and magnetic or electric fields on the hydrogen‐bond environment. For P. vannamei, ultrasound‐assisted high‐pressure freezing (US‐HPAF) lowered drip from 7% to 4% beyond either method alone (Hu et al. 2022). PEF‐ABF in S. salar approached PSF refinement at one‐fifth the energy cost (Rondineli and Silva 2024). However, additive versus synergistic effects are seldom formally tested through factorial designs with statistical interaction terms; reported “synergies” may compound independent positive effects rather than represent true cooperative enhancement (Hu et al. 2022). A more cautious reading is that hybrid approaches mostly broaden the operating window rather than raise peak performance: A UAF protocol tuned for shrimp may fail for cod, whereas a US‐HPAF protocol at intermediate settings tolerates a wider range of species.

4.7. Cross‐Technology Mechanistic Synthesis

Three decisive factors determine outcomes across archetypes (Figure 2; Tables 1 and 2): They are the supercooling degree at the start of ice nucleation, the speed of latent‐heat removal in the later crystal growth stage, and the influence degree exerted on natural protein and lipid structures. HPAF/PSF excels at the first, UAF at the second, and MWAF/RFAF at the third. Each technology fails when pushed beyond the parameter window where its dominant mechanism is operative: Pressure beyond 250 MPa denatures myosin, ultrasound above 0.4–0.5 W/cm² in continuous mode, and microwave at high‐duty cycle erodes the supercooling it is meant to maintain. Cross‐species analysis identifies intramuscular lipid content, collagen cross‐linking, and cuticle architecture as the binding boundary conditions that translate the same physical mechanism into different quality outcomes. Section 5 develops this into a species‐to‐strategy decision tree, as an alternative to the implicit assumption that a single protocol can serve all aquatic species. Thus, Figure 2 should be read as a stage‐specific mechanism map rather than as a general technology summary. HPAF/PSF mainly regulates nucleation thermodynamics through pressure‐induced supercooling and rapid pressure release, whereas UAF affects nucleation and early crystal growth through cavitation, microstreaming, and secondary crystal fragmentation. RFAF/MWAF acts mainly through dielectric regulation of the unfrozen interfacial phase and local temperature oscillation near the freezing front. By contrast, MFAF and low‐field EFAF are placed in a lower confidence category because their proposed effects on water structure, hydrogen‐bond networks, or nucleation barriers remain difficult to separate from indirect effects such as vibration, eddy‐current heating, or convective disturbance. Therefore, when the field intensity exceeds the usable process window, quality regulation may shift into molecular damage, including protein denaturation, lipid oxidation, thermal stress, or poorly reproducible field effects.

5. Integrated Comparative Framework, Critical Appraisal, and Research Agenda

5.1. A Three‐Tier Framework: Mechanism–Process Window–Engineering Envelope

Section 4 establishes that physical‐field‐assisted freezing is best understood not as a single emerging technology but as a portfolio of mechanistically distinct interventions. To translate this into operational guidance, we propose a three‐tier framework (Figure 3). Tier 1 (mechanism) classifies each technology by its dominant physical character—chemical‐potential shift (HPAF/PSF), nucleation multiplication (UAF, hybrid US‐HPAF), interfacial regulation (RFAF/MWAF), or unverified mechanism (MFAF, low‐field EFAF). Tier 2 (process window) defines the parameter range within which the dominant mechanism produces net positive outcomes without crossing the protein‐denaturation, lipid‐oxidation, or interfacial‐fracture thresholds (Tables 1 and 2). Tier 3 (engineering envelope) maps capital cost, energy demand, throughput geometry, food‐grade scale‐up status, and regulatory acceptance. This evaluation system sets clear practical standards: only techniques that satisfy all three practical requirements can be promoted for industrial use. At present, only high‐pressure‐assisted freezing works well for high‐grade aquatic products, whereas continuous UAF is suitable for quick‐frozen shellfish and thin fillets under 30 mm in thickness. No other methods have fully reached this practical standard yet. This restrictive reading is at odds with the more enthusiastic framing of recent reviews (Peng et al. 2025; Pu et al. 2025) but is consistent with the scarcity of commercial deployments outside Japan, the EU, and a small number of Chinese facilities (Pou 2021). The key conclusion from Figure 3 is therefore restrictive: A physical‐field freezing technology is practically meaningful only when mechanism, process window, and engineering feasibility are satisfied simultaneously. Isolated positive laboratory results, especially those based only on crystal size or a single quality index, are insufficient to justify industrial application.

FIGURE 3.

A three‐tier framework for evaluating physical‐field‐assisted freezing of aquatic products. EFAF, electric field‐assisted freezing; HPAF, high‐pressure‐assisted freezing; MFAF, magnetic field‐assisted freezing; MWAF, microwave‐assisted freezing; PSF, pressure‐shift freezing; RFAF, radiofrequency‐assisted freezing; UAF, ultrasound‐assisted freezing; US‐HPAF, ultrasound‐assisted high‐pressure freezing.

5.2. Species‐to‐Strategy Decision Tree

Building on the four‐category stratification of Section 2.4 and the lipid, collagen, and cuticle constraints identified in Section 4.7, Figure 4 synthesizes the species‐to‐strategy mapping in four panels: archetype‐level sensitivity scoring (A), cross‐species technology suitability (B), recommended primary/alternative interventions and target outcomes (C), and the operational decision tree (D). Lean gadoids and percoids (Gadus morhua, Atlantic cod; Pollachius virens, saithe; and Lateolabrax japonicus, Japanese sea bass) in fillet form are best served by single‐ or dual‐frequency UAF at 25–30 kHz, 0.2–0.3 W/cm², treatment time scaled to fillet thickness (5 min for ≤25 mm; 8 min for 25–40 mm). UAF in this group addresses the texture‐loss pathway by refining ice–crystal size and limiting myofibrillar disruption, but it does not directly inhibit the TMAO demethylase activity that generates formaldehyde and underlies the “spongy” toughening identified in Section 2.4 as the rate‐limiting defect of this category. Consistent with the conclusion in Section 2.4 that bleeding quality and dark‐muscle removal are at least as critical as the freezing step itself, the UAF protocol should therefore be paired with thorough bleeding, deboning, and removal of dark muscle and viscera at primary processing—these upstream interventions, not the freezing technology, are the operative controls on formaldehyde generation (Li, Wang, Yanagita, et al. 2024). Fatty pelagic species (S. salar; Thunnus thynnus, Atlantic bluefin tuna; and Scomber japonicus, chub mackerel) are oxidation‐limited rather than crystal‐limited: PUFA oxidation, not ice–crystal damage, is the rate‐determining failure mode identified in Section 2.4. The primary intervention is therefore exclusion of oxygen and light combined with storage below −30°C—vacuum packaging (or modified‐atmosphere packaging with oxygen scavengers where feasible) is the direct control on phospholipid radical chain propagation and on the EPA/DHA‐driven cosmetic yellowing of salmon flesh (Russo et al. 2024; Shi et al. 2022). PSF at 150–200 MPa with rapid release is recommended as a complementary physical intervention that refines ice–crystal morphology, preserves the myofibrillar lattice, and reduces sarcolemmal rupture, thereby attenuating the release of heme iron and lipoxygenases that otherwise accelerate the oxidative cascade; PSF itself does not suppress lipid oxidation and, above 200 MPa, may even promote it through pressure‐induced perturbation of phospholipid bilayers (Russo et al. 2024). The causal hierarchy in this group is, therefore, packaging and storage temperature first, physical‐field freezing second. Crustaceans (L. vannamei, Procambarus clarkii, red swamp crayfish; Chionoecetes opilio, snow crab; and P. trituberculatus) respond best to PSF at 100–150 MPa or to LNF at −90°C to −100°C, where the moderate cryogenic temperature avoids cuticle fracture while suppressing polyphenol‐oxidase activation and melanosis (Hashempour‐Baltork et al. 2024; Yan et al. 2023). Its working principle is straightforward. Fast latent‐heat removal can reduce ice crystal size and also cut down the heating duration that triggers PPO and serine protease activity. These two enzymes are confirmed earlier as the main causes of darkening and texture deterioration. For squid, octopus, and cuttlefish, suitable high‐pressure‐assisted freezing should be controlled steadily between 150 and 180 MPa. This range must stay below 200 MPa, because higher pressure is proven to damage their skin and connective tissue collagen structure (Nilsuwan et al. 2024; Zura‐Bravo et al. 2025). These aquatic species are highly sensitive to elevated pressure levels. When processing pressure reaches 200 MPa and above, the treatment no longer optimizes ice crystal formation, and excessive pressure will directly induce collagen denaturation, making the freezing treatment ineffective. When applied within 150–180 MPa, this method efficiently targets the core problem mentioned earlier, namely, collagen aggregation along growing ice crystals. It can reduce ice crystal size, cut down the contact area between ice crystals and collagen, and keep the original spatial structure of Types I and V collagen intact (Nilsuwan et al. 2024; Sadot et al. 2017). Surimi and gel‐type products are uniquely served by HPAF, where pressure‐induced gel densification compounds crystal‐refinement benefit (Liu, Zeng, et al. 2025); For surimi products, combining high‐pressure‐assisted freezing with antifreeze peptides extracted from red sea bream, obscure puffer skin, or silver carp scales delivers better preservation effects. This approach also requires far lower additive dosage compared with the widely used formula containing 4% sucrose and 4% sorbitol (Chen et al. 2022; Fan et al. 2024; Yang, Jiang, et al. 2023). The key conclusion from Figure 4 is that technology selection should begin with the dominant failure mode of each aquatic‐product category rather than with the availability of a freezing device. Lean fish are mainly constrained by formaldehyde‐related textural toughening, fatty fish by lipid oxidation, crustaceans by melanosis and protease‐driven softening, cephalopods by collagen aggregation, and surimi‐type products by the preservation of gel‐forming protein functionality. Therefore, these product categories require different intervention strategies, and a single physical‐field protocol cannot be assumed to be universally applicable.

FIGURE 4.

Species‐specific freezing sensitivity and physical‐field integration strategy for aquatic products. (A) Species Archetypes and Intrinsic Freezing Sensitivities. (B) Species Technology Suitability Matrix. (C) Recommended Physical‐Field by Species. (D) Decision Tree for Species‐to‐Strategy Selection. DHA, docosahexaenoic acid; EFAF, electric field‐assisted freezing; HPAF, high‐pressure‐assisted freezing; MFAF, magnetic field‐assisted freezing; MWAF, microwave‐assisted freezing; PSF, pressure‐shift freezing; PSF, pressure‐shift freezing; RFAF, radiofrequency‐assisted freezing; UAF, ultrasound‐assisted freezing.

The species‐to‐strategy framework is therefore restrictive rather than merely classificatory. A technology should not be recommended simply because it improves one physical index under laboratory conditions. It should be considered suitable only when its dominant mechanism targets the main failure pathway of the product. For example, crystal refinement alone cannot solve formaldehyde‐related toughening in lean fish, oxygen‐driven oxidation in fatty fish, melanosis in crustaceans, or collagen aggregation in cephalopods. This interpretation narrows the practical scope of each technology and avoids the overgeneralized conclusion that a single physical‐field protocol can be applied to all aquatic products.

5.3. Standardized Reporting Matrix

The biggest hurdle hindering comparative research lies in inconsistent experimental data recording standards. Ultrasound studies frequently omit power density and duty cycle; magnetic‐field studies rarely report field‐strength uniformity; and pressure studies often quote nominal vessel pressure without temperature–pressure trajectory. We therefore propose the minimum reporting matrix in Table 3, covering field‐specific parameters, sample geometry (binomial nomenclature, dimensions, and lipid/collagen content), freezing kinetics (medium, surface heat‐transfer coefficient, and characteristic freezing time from −1°C to −7°C), storage conditions (T′_g‐relative storage temperature, fluctuation amplitude), and quality endpoints (drip loss, ice–crystal size with measurement method declared, TBARS, carbonyl content, protein‐secondary‐structure indices, and sensory scores). Consistent reporting along these lines would make published studies comparable and open the way to the meta‐analyses that are currently not feasible (Pérez‐Bermúdez et al. 2023; Jia, Chen, et al. 2022; Yang, Zhang, et al. 2025).

TABLE 3.

Three‐axis evaluation: technology readiness level (TRL), industrial feasibility, sustainability, and sensory profile.

Technology	TRL a	CapEx b for 500 kg/h ($)	Energy increment c (kWh/kg)	Throughput	Commercial deployment d	Sensory consensus	Net evidence rating	References
HPAF/PSF	6–7	1.5–3 M	0.4–0.6	Batch (50–100 L)	Premium oysters/sashimi (JP, EU)	Texture superior; opacity at >250 MPa	Strong (positive)	Roobab et al. (2022)
UAF (continuous)	6–7	0.10–0.25 M	0.05–0.10	Continuous ≥500 kg/h	Shrimp, shellfish (CN, VN)	Indistinguishable from fresh ≤60 days	Strong (positive)	Cheng et al. (2017)
MUIF (multi‐freq)	5–6	0.20–0.40 M	0.08–0.12	Continuous	Pilot (CN)	Comparable to fresh at 175 W	Moderate–strong	Ma et al. (2021)
HPT	5–6	1.0–2 M	0.2–0.3	Batch	Limited high‐value tuna	Surface denat. risk	Moderate	Roobab et al. (2022)
LNF	7–8	0.30–0.80 M	0.5–1.5	Continuous (IQF)	Cryogenic IQF (global)	Surface‐fracture risk at <−110°C	Strong (positive)	Huang et al. (2024)
PEF + ABF	5	0.20–0.40 M	0.10–0.15	Continuous	Pilot EU salmon	Limited audit	Moderate	Peng et al. (2025)
HVEF (thaw)	5–6	0.15–0.30 M	0.10–0.20	Semi‐cont.	Pilot tilapia	Slight oxidative penalty	Moderate	Çalışkan Koç et al. (2025)
RFAF	4	0.15–0.30 M	0.15–0.25	Semi‐cont.	None	None published	Low–moderate	Hafezparast‐Moadab et al. (2018)
MFAF (CAS‐type)	3–4	0.20–0.50 M	0.05–0.08	Batch	Vendor‐promoted	None reproducible	Low	Otero et al. (2016)
EFAF (active)	3	—	0.05	Batch	None	None published	Low	Sun et al. (2024)
MWAF	3	—	0.20–0.35	Pilot	None	None	Low	Anese et al. (2012)
US‐HPAF hybrid	3–4	2.5–4 M	0.45–0.65	Batch	Lab/Pilot	Limited audit	Moderate	Hu et al. (2022)

Abbreviations: ABF, air‐blast freezing; EFAF, electric field‐assisted freezing; HPAF, high‐pressure‐assisted freezing; HPT, high‐pressure thawing; HVEF, high‐voltage electrostatic field; LNF, liquid‐nitrogen freezing; MFAF, microwave‐assisted freezing; MUIF, multifrequency ultrasound‐assisted immersion freezing; MWAF, Microwave‐assisted freezing; PSF, pressure‐shift freezing; RFAF, radiofrequency‐assisted freezing; UAF, ultrasound‐assisted freezing; US–HPAF, ultrasound‐assisted high‐pressure freezing.

^a TRL (technology readiness level) was assigned according to the maturity of food‐grade equipment and the level of validation in aquatic‐product systems. TRL 1–3 cover basic research to preliminary laboratory verification; TRL 4–5 indicates laboratory or pilot‐scale validation; TRL 6–7 indicate prototype or near‐commercial demonstration in relevant food‐processing environments; and TRL 8 indicates established commercial use in related frozen‐food operations.

^b CapEx values are approximate order‐of‐magnitude estimates for a 500 kg/h processing capacity and are intended for comparison rather than quotation.

^c Energy increment refers to additional energy demand above the baseline freezing process and does not include upstream raw‐material handling, packaging, frozen storage, or distribution.

^d Commercial deployment was classified from reported pilot‐scale or industrial applications, rather than from laboratory proof‐of‐concept alone.

5.4. Methodological Limitations and Reproducibility

Several systemic weaknesses compromise current evidence. Ice–crystal sizing relies predominantly on freeze‐substitution and freeze‐drying followed by light microscopy, both of which introduce known artifacts (solvent‐mediated swelling; secondary recrystallization during sectioning); cryo‐SEM and x‐ray micro‐CT, less artifact‐prone, are deployed in <15% of published aquatic studies (Pérez‐Bermúdez et al. 2023). Many physical‐field studies use a single biological replicate or batch, conflating inter‐batch variability with treatment effect (Cheng et al. 2017). Statistical power is rarely justified, and effect sizes for marginal benefits (e.g., MFAF at <100 mT) are typically smaller than the inter‐batch variance reported in conventional‐freezing studies (Otero et al. 2018). Hybrid configurations are particularly under‐reported: Factorial designs with formal statistical interaction tests are exceptional rather than routine (Hu et al. 2022). Meta‐analytic synthesis is currently impossible because of these limitations, and the immediate priority is methodological harmonization rather than further single‐laboratory studies.

5.5. Sustainability, Cost, and Sensory Acceptance

Environmental and economic considerations have been understated in prior reviews. UAF and PEF pretreatment appear to offer the most favorable energy profile, with marginal increments of 0.05–0.15 kWh/kg over baseline freezing and capital costs below $250,000 for tunnel‐scale installations (Cheng et al. 2017; Rondineli and Silva 2024). HPAF/PSF carries 2–3‐fold higher capital cost and comparable operating energy, but shelf‐life extension and reduced product loss may partly offset downstream cold‐chain burdens in selected high‐value products. However, any net carbon benefit remains hypothetical and should be verified by process‐specific life‐cycle or carbon‐footprint assessment rather than inferred from equipment energy alone (FAO 2024; Roobab et al. 2022; Iamchamnan et al. 2025). MFAF and EFAF, despite low operating energy, offer too little quality benefit to offset capital cost at present.

Sensory data are the most fragmentary part of the evidence base. Trained‐panel and consumer‐triangle studies have been conducted for PSF‐treated S. salar (superior texture and freshness vs. air‐blast) and MUIF‐treated L. crocea (indistinguishable from fresh after 60 days) (Bian et al. 2022; Ma et al. 2021). High‐pressure‐induced opacity in red‐fleshed species above 250 MPa represents a hard sensory ceiling (Russo et al. 2024). Sensory data for MFAF, EFAF, RFAF, and MWAF in aquatic products remain too sparse for quantitative inference—a gap of obvious commercial relevance that the next wave of multi‐laboratory studies must address.

5.6. Prioritized Research Agenda and Future Outlook

Five research priorities emerge. First, mechanistic resolution of MFAF and EFAF in aquatic muscle through in situ NMR and neutron‐scattering measurements during freezing under controlled fields to establish whether claimed effects exist at industrially accessible field strengths. Second, predictive freezing models that incorporate the visco‐elastic resistance of the protein matrix, anisotropic heat conduction along fiber bundles, and explicit T′_g constraints, replacing the dominant Stefan‐problem idealization (Huo et al. 2024). Third, systematic factorial dose–response mapping for hybrid configurations, with formal interaction testing to distinguish synergy from additivity. Fourth, scale‐up engineering for UAF and HPAF to whole‐fish and pallet‐scale geometries, addressing acoustic‐attenuation and pressure‐vessel‐volume bottlenecks. Fifth, longitudinal studies tracking cold‐chain stability beyond 90 days under realistic ±2–3°C fluctuations, coupling crystal evolution with molecular oxidation markers, including ferroptosis‐related lipid peroxidation (Huang et al. 2024).

Three converging trends shape the longer term outlook. Real‐time process‐analytical technology (dielectric, ultrasonic, NMR, hyperspectral, and electronic‐nose sensors integrated into freezing tunnels) will allow closed‐loop adjustment of field parameters in response to nucleation events, replacing current open‐loop dosing (Jia, Chen, et al. 2022; Wang, Liu, et al. 2024). Coupling with high‐throughput screening of cryoprotective peptides, antifreeze proteins, and natural deep‐eutectic‐solvent‐based formulations is moving the most promising hybrid technologies from physical‐only to integrated physical–chemical regulation (Fan et al. 2024; Jiang et al. 2026; Chen et al. 2022; Castro et al. 2018). The existing industrial use of high‐pressure processing in seafood, including microbial control, shellfish shucking and, clean‐label ready‐to‐eat products, suggests a possible commercialization basis for pressure‐based freezing in selected high‐value aquatic products (Pou 2021; Roobab et al. 2022). However, the application of HPAF/PSF to freezing should be regarded as a future possibility rather than a firm 5–10 year prediction, because freezing‐specific validation, equipment throughput and cost barriers remain unresolved. Coupled with AI‐driven freezing‐process optimization now emerging in meat and dairy (Xia et al. 2026), the next generation of aquatic‐product freezing platforms may become hybrid, sensor‐integrated, and species‐adaptive. Progress in mechanistic understanding, in reporting standards, and in engineering scale‐up will need to proceed together; advances confined to any one of these are unlikely to be sufficient on their own.

6. Discussion and Conclusions

Several conclusions follow from the evidence reviewed here, and they qualify rather than confirm the optimistic tone of much of the recent literature on physical‐field‐assisted freezing of aquatic products. We summarize them below rather than reiterate the discussion of each technology.

It is widely believed that smaller ice crystals always bring better product quality, yet this rule only works within a certain range. Excessively refined ice crystals will trigger other forms of tissue damage. Ultra‐low temperature IF can form ice crystals smaller than 20 µm, but severe thermal stress will crack the outer shell of shrimp and crayfish. Fast freezing without cryoprotectants creates a large ice–water interface area, which easily induces protein structural changes on tissue surfaces. Recent studies also confirm that deep freezing may trigger lipid peroxidation linked to cell ferroptosis, a kind of quality loss that cannot be judged simply by physical tissue indicators. In fact, ice crystal size is only one adjustable factor in the complex physical and chemical system. Researchers need to comprehensively optimize multiple indicators, including interface area, unfrozen concentrated solution state, glass transition temperature, protein stability, and free radical generation, instead of only focusing on crystal size. This research idea can also resolve many conflicting conclusions in existing studies, such as inconsistent results regarding ultrasound power and hydrostatic pressure effects.

Different aquatic species have inherent differences that cannot be ignored in actual production. Lean white fish, high‐fat pelagic fish, crustaceans, and cephalopods all suffer quality deterioration through completely different pathways. Specifically, lean fish turn tough due to formaldehyde accumulation caused by TMAO decomposition, fatty fish go bad mainly from polyunsaturated fatty acid oxidation, crustaceans darken owing to polyphenol oxidase activity, whereas cephalopods deteriorate as collagen gathers around growing ice crystals. A freezing method effective for one type of aquatic product may have no positive effect or even cause adverse results on others. For example, magnetic field treatment used to prevent shrimp blackening cannot inhibit formaldehyde production in cod, and ultrasound parameters suitable for maintaining tuna lipid stability fail to slow collagen deterioration in octopus. Therefore, selecting targeted processing methods according to species characteristics is not just an auxiliary classification method, but a practical guiding principle that fits actual production needs.

The functional mechanism of magnetic field assisted‐freezing and low‐frequency electric field‐assisted freezing still needs more rigorous verification. Theoretical calculation results show that the impact of common food‐grade magnetic field equipment on ice nucleation is far weaker than the actual nucleation energy barrier. The 1–3°C supercooling effect observed in many tests can also be achieved by traditional well‐controlled freezing equipment. Most existing experimental data have two obvious flaws: The treatment effect is even smaller than normal sample batch differences, and experimental results obtained by different magnetic field devices are hard to repeat. Without real‐time in situ detection data, such as neutron scattering and nuclear magnetic resonance during freezing, most repeatable experimental phenomena are more likely caused by mechanical vibration, eddy current heat, and fluid flow changes rather than direct interaction between physical fields and water molecules. Blindly affirming the effect of such physical‐field freezing will lead to research bias, where positive experimental results keep increasing, whereas core mechanism problems remain unsolved.

The core bottleneck restricting the development of this field has shifted from basic mechanism exploration to unified experimental standards and industrial scale‐up. Disordered experimental settings make it impossible to summarize regular research conclusions. Many ultrasound freezing tests ignore power density parameters, pressure freezing tests lack complete temperature and pressure change data, and magnetic field‐related experiments do not record field uniformity. These incomplete data cannot be used for comprehensive comparative analysis. Besides, many single‐batch experiments fail to set reasonable statistical standards, confusing sample differences with actual treatment effects, which is more likely to cause wrong conclusions when the overall treatment effect is weak. Unifying standardized experimental reporting rules can effectively sort out existing research data and form a complete and comparable research system. In addition, large‐scale industrial equipment has clear practical limitations, such as limited high‐pressure container volume, limited effective penetration depth of ultrasound, and difficult matching of radio frequency equipment in assembly line production. These practical problems are often ignored in laboratory research, which also explains why relevant research papers are numerous, whereas actual industrial application cases are few.

Glass transition temperature is also an easily overlooked key index for frozen product storage. Most cold‐chain logistics maintain the storage temperature at −18°C, which is only 5–8°C lower than the glass transition temperature of ordinary fish muscle. Although this temperature is theoretically able to slow down tissue changes, the temperature gap is too narrow to completely stop internal substance movement. Subtle molecular relaxation on ice crystal surfaces and frequent small temperature fluctuations in cold‐chain transportation will continuously promote ice crystal recrystallization and protein aggregation, leading to slow and inevitable quality decline during storage. Optimizing the freezing process cannot make up for the defect of insufficiently low storage temperature. For high‐value aquatic products, properly lowering the storage temperature by 7–10°C can achieve better fresh‐keeping effects. This adjustment is technically feasible, and its popularization is only restricted by cost and traditional industry habits. It can be seen that reasonable post‐freezing storage conditions are equally important as advanced freezing processing technology.

In the future development trend, various physical‐field freezing technologies will no longer develop independently, and integrated physical and chemical combined regulation modes will become the mainstream. Natural antifreeze active substances extracted from aquatic organisms and compound cryoprotectants can regulate water molecular state, hydrogen‐bond arrangement, and protein stability through chemical pathways that share the same action targets as physical‐field freezing methods. Existing experimental data prove that combined treatment can achieve ideal fresh‐keeping effects under lower physical‐field intensity, which not only reduces equipment operation difficulty but also lowers the risk of protein denaturation. As for the synergistic effect of combined technologies, most observed results are still within the range of simple superposition effects. Its greatest practical value lies in reducing the sensitivity of processing parameters and improving the stability of fresh‐keeping effects, which is more in line with commercial production demands than pursuing extreme fresh‐keeping performance, especially for raw materials with unstable quality between batches.

From the perspective of actual industrial promotion, high‐pressure‐assisted freezing is already widely used in high‐value fish fillets and surimi products, and continuous ultrasound freezing technology has also been applied in large‐scale quick freezing of shellfish and thin fish fillets. Most other physical‐field freezing technologies are still stuck in the laboratory research stage. With the growing market demand for aquatic products and the persistent 10%–25% quality loss rate in cold‐chain circulation, the industry still has huge development space. Yet simply relying on physical freezing principles cannot solve all practical problems. The follow‐up research needs to focus on improving experimental repeatability, unifying test standards, carrying out long‐term actual cold‐chain storage verification, and breaking through various equipment application limitations ignored in laboratory research. With the gradual maturity of real‐time monitoring technology, intelligent parameter adjustment system, and green cryoprotectant technology, a new generation of intelligent freezing equipment integrating multiple functions, precise sensing, and species‐adaptive regulation will be developed. The popularization speed of such new equipment depends not on discovering new physical principles, but on the coordinated progress of mechanism research, experimental standards, and industrial application technology. Single‐sided breakthrough in any single field cannot make up for the deficiencies in other research directions.

Author Contributions

Yuyang Zhang: writing – original draft, writing – review and editing, visualization, methodology, investigation. Ting Xiao: writing – review and editing, resources, methodology. Maninder Meenu: visualization. Xinxin Li: visualization. Tao Song: writing – review and editing, methodology. Sinan Zhang: investigation, visualization. Yuxiao Mao: investigation, software. Lihui Hu: methodology, resources. Ying Liu: methodology, formal analysis, visualization. Hosahalli S. Ramaswamy: methodology, formal analysis, visualization. Yong Yu: funding acquisition, conceptualization, project administration, supervision.

Funding

This work was supported by the National Natural Science Foundation of China (grant number 31871892).

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

Data will be made available on request.