Liquid immersion cooling combines an electrically insulating, thermally conductive fluid with power-conversion hardware, targeting the thermal and voltage stresses created by wide-band-gap devices, high-frequency switching and tightly integrated assemblies [S1][S2].
The approach applies to utility-scale transformers, high-voltage rectifiers, data-center equipment, DC/DC converters and high-power electronic chips, with mineral oil and FR3 fluids identified as candidate dielectric coolants [S1][S3][S4].
Cooling architecture and operating principle
Liquid immersion places the electronic assembly in a nonconductive liquid so the coolant removes heat directly from exposed components and carries it to a heat exchanger, rather than relying on air flowing across a heat sink [S1][S4].
Transformer and high-voltage-rectifier applications have more than 100 years of liquid-immersion history, while data-center use extends beyond 20 years [S1]. This history supports the basic concept, but the materials, packaging and electrical clearances required by modern wide-band-gap power converters remain design constraints rather than a plug-and-play replacement for air cooling [S1].
Immersion also removes the need for a conventional fan-and-sink path in principle, although pumps, reservoirs, heat exchangers, seals, controls and fluid-property monitoring still form part of a practical plant [S3][S4]. The engineering objective is lower thermal resistance, smaller heat-transfer surfaces and more uniform cooling across densely populated power boards [S2][S5].
Transformer, rectifier and data-center precedent
Liquid-immersed transformers provide a mature reference point for utility equipment, while high-voltage rectifiers extend the same thermal principle toward power conversion [S1]. A conventional power transformer selection remains tied to insulation coordination, dielectric strength, thermal design, mechanical support and maintainability, so the fluid alone does not determine transformer suitability [S1].
Data-center equipment supplies a second precedent through immersion-cooled servers and power infrastructure, where heat removal must be continuous and compatible with dense electrical layouts [S1]. In those installations, the cooling loop must coordinate fluid temperature, heat-exchanger capacity, pump behavior and equipment availability rather than treating the coolant as a passive bath [S4].
Utility-scale equipment and data-center systems differ in duty cycle, enclosure size, service access and acceptable dielectric fluids, but both show that immersion is not a laboratory-only concept [S1]. For a new build, transformer thermal requirements should therefore be separated from converter-board dielectric and maintenance requirements, with qualified fluid and materials specified for each assembly [S1].
FR3 and mineral-oil dielectric fluids
FR3 was evaluated as a renewable alternative to mineral oil and is described as soy-based, non-toxic and biodegradable, giving it a materially different environmental profile from conventional petroleum-derived mineral oil [S1].
Sandia’s tests compared air and Cargill FR3 across conductor clearances from 0.1 mm to 4 mm [S1]. In air, breakdown occurred at 2.1 kV for 0.1 mm, 3.3 kV for 1 mm and 5.5 kV for 4 mm under the reported HVDC test condition, while the corresponding HVAC results were 1.5 kV, 1.9 kV and 3.7 kV; no breakdown occurred in FR3 in the reported test matrix [S1].
Those results are test evidence for a specific fluid, board construction and geometry, not a universal clearance table for every converter. PCB design remains dictated by creepage, clearance and electric-field intensity, with a 2D electrostatic model examining conductor shape and spacing under 3 kV excitation [S1].
Single-phase and two-phase immersion
Single-phase immersion keeps the dielectric liquid in a liquid state around the electronics, while two-phase immersion uses a fluid that changes state and transfers heat through latent heat during boiling and condensation [S3][S5]. The two approaches are also distinguished from liquid cold plates, which circulate coolant through channels against a separate heat-transfer surface rather than submerging the active assembly [S5].
The decision is not simply liquid versus air. Air cooling is familiar and serviceable but depends on fin area, airflow and thermal resistance; liquid cold plates support single-phase aqueous or oil flow and two-phase dielectric-fluid or refrigerant operation; full immersion places the board or device directly in the dielectric environment [S5].
Two-phase systems can raise heat transport capacity because evaporation adds latent heat to convective heat transfer, but they add vapor handling, condensation control and fluid-management requirements [S3]. Single-phase immersion is generally simpler in its phase-control architecture, so it is a credible starting point when the required heat flux can be met with liquid convection and conduction [S3][S5].
Device temperature and power-density limits
Power-device packaging and adjacent materials limit practical silicon junction temperatures to about 175 °C, even though the semiconductor itself may tolerate higher temperatures, while the maximum safe junction temperature of SiC devices can exceed 300 °C [S5]. Removing heat at the device level therefore has direct value where junction temperature controls reliability, efficiency and volumetric power density.
Wide-band-gap semiconductors such as SiC and GaN support high-frequency, high-power and fast-switching operation, but that performance raises electric-field stress and heat-flux density [S1]. Sandia’s early device example used a 3.3 kV, 50 A device with approximately 125 W dissipation, and its work integrated control, sensing, communication and power circuitry on PCBs and other packages for kilowatt-class throughput [S1].
Cooling selection must remain tied to the complete thermal path. An immersion fluid that limits semiconductor temperature does not remove constraints from solder, wire bonds, dielectric laminates, encapsulants, connectors or PCB insulation [S1][S5]. Specifiers should use measured device and board temperature limits, not only coolant temperature, as the acceptance criteria.
Design selection for compact power boards
Immersion is most attractive for high heat-flux, space-constrained power conversion equipment, including DC/DC converters, high-power chips, rectifiers and compact medium-frequency systems [S3][S4]. It is less compelling where airflow already handles the thermal duty with little loss of enclosure space, where a low-cost repair demands simple dry servicing, or where the fluid cannot be qualified against every material in the assembly [S2][S5].
For a compact immersed PCB, the selection sequence should cover fluid dielectric strength, thermal conductivity and operating-temperature range, compatibility with PCB laminates, solder, coatings, connectors and encapsulants, then component spacing, partial-discharge behavior, leakage and maintenance access [S1][S2]. Electric-field modelling should be repeated for the final conductor geometry because pad shape and spacing can alter field concentration at 3 kV excitation [S1].
Equipment built around a dry-type transformer or conventional ventilated enclosure should not be converted to immersion by fluid substitution alone, because the transformer dielectric system, enclosure, bushings and protection arrangement have different requirements from an immersed converter board [S1].
Verification, maintenance and failure controls
A credible immersion design requires testing beyond a single thermal run: dielectric withstand, partial discharge, leakage current, temperature rise, fluid aging and material compatibility should be verified on representative assemblies [S1][S2]. Sandia evaluated FR3 dielectric properties against air at 0.1 mm, 1 mm and 4 mm clearances, while the thermal model used a simplified 1 cm by 1 cm flat plate [S1].
The flat-plate model is useful for comparing heat-transfer behavior, but it does not reproduce every component, lead, void, PCB stack-up or winding in a production converter. Validation should therefore move from fluid screening to complete assemblies, with sensors placed at semiconductor cases, magnetics, bushings or power terminals, not only in the bulk fluid [S1][S2].
Common failure modes to control include dielectric breakdown, partial discharge, corrosion or incompatibility, fluid leakage, blocked heat exchange, pump failure, local hot spots and uneven temperature distribution [S2][S3]. An alarm and shutdown strategy should therefore cover coolant temperature, tank level, pump state and heat-exchanger performance, while also protecting exposed conductors under the applicable insulation-coordination requirements [S1].
Market and engineering trend
The engineering trend is a move from air-cooled modules toward integrated liquid paths, with direct immersion emerging as one route for high-density converters rather than a universal standard [S2][S5]. For data-center power chains, that trend links directly to DC microgrids in industrial plants, where conversion density, redundancy and thermal capacity must be balanced.
Mineral oil carries the benefit of long operating history in transformers and high-voltage rectifiers, while FR3 introduces a renewable, soy-based dielectric alternative with the reported non-toxic and biodegradable profile [S1]. The decisive issue is qualification: fluid chemistry, geometry, voltage level, heat load and service conditions determine whether a candidate passes [S1][S2].
Trackable engineering signals include wider use of dielectric immersion in medium-frequency converters, published breakdown data for final PCB geometries, and field validation of FR3 or other candidate fluids at utility voltage levels [S1][S3]. Procurement teams should request fluid-property data, test reports and material-compatibility evidence before comparing an immersed converter with an air-cooled or cold-plate alternative [S1][S2].