KEY CONCEPTS
• Air cooling is no longer effective for our increasingly powerful computer chips and circuit boards in crowded computer bank designs.
• More efficient liquid cooling systems are being developed where entire circuit boards and server landscapes are immersed in cooling oils that can more effectively dissipate heat.
• Evaluation of the cooling fluid performance and its compatibility with the materials it is cooling is carried out in the test laboratory using accelerated aging procedures.
• Test standards for data center applications are not yet known, but work is occurring to develop the appropriate standards.

Digitalization is part of our everyday lives with global demand only increasing for information, networking, cloud service, storage applications and the associated digital infrastructure. Increasingly powerful computer chips and circuit boards require effective cooling. Because of this, older air cooling systems are gradually being replaced by more efficient liquid cooling systems. These newer cooling systems include immersion cooling where entire circuit boards and server landscapes are immersed in cooling oils that can effectively dissipate heat, as well as direct-to-chip cooling where components are indirectly cooled. 

With single phase immersion cooling, the circuit board and server materials must be compatible with the immersion fluids. While the fluids for immersion cooling generally have to be dielectric, indirect cooling does not usually require the liquids to be dielectric. However, there are certain requirements here, too, which can also include conductivity. The development of new fluid formulations for immersion cooling purposes is based on static immersion tests performed on a laboratory scale of relevant materials in the experimental oil. 

This article focuses on the development and test criteria for novel heat transfer oil formulations in static immersion tests and addresses issues of material selection and test parameters. The evaluation of the oil performance and the compatibility of the materials is carried out in the test laboratory using procedures that simulate accelerated aging. The mechanical material characteristics are determined before and after storage at elevated temperatures of the test oil. However, there are only a limited number of test criteria for data center applications like fluid requirements, which makes it difficult to find generally acceptable tests. There is still much to be done.

Test specifications
STLE member Johannes Fischer is head of laboratory media compatibility oil/lubricants/fluids at Minz Prüf + Test GmbH, an accredited independent testing laboratory for plastics and elastomers in Limburg an der Lahn, Germany. Fischer says that defined test standards are often used for static laboratory tests, such as ISO 1817 or ASTM D471.^1,2 These standards have existed for a long time and are mainly used to assess the compatibility of oils with sealing materials such as O-rings. In general, Fischer says these tests “compare the hardness, tensile strength and elongation at break both before and after storage in the fluid.” The materials are therefore a useful aid for classifying fluids. Several reference elastomers with a known composition are often used as these materials. However, application-specific materials used in practice are also widely applied in compatibility testing. 

Nevertheless, Fischer notes that these standards clearly reach their limits with new types of immersion fluids for battery cooling and data centers and cannot be directly applied to the new cooling oils. The test conditions, which already deviate from reality (i.e., static test versus dynamic relativity), are made even more challenging here. In particular, heat flows with constant fluid movement due to pumping and other causes are not taken into account. Other factors to be considered when addressing the testing of new immersion coolants are temperature cycles, electric fields, long-term-permeation or the fluid itself aging. This requires workarounds such as simulating temperature gradients with climate chambers with corresponding heat ramps if necessary. 

Fischer observes that many insulating materials (e.g., polyvinyl chloride [PVC], polyethylene [PE] and thermoplastic elastomer [TPE]) react sensitively to additive extraction (i.e., loss of plasticizer) and show a change in dielectric strength. This means that electrochemical side reactions occurring on the sealing surfaces or insulators can be tested on the materials after aging by measuring the dielectric strength or surface resistivity.

Fischer notes that these new fluids also operate at a lower temperature range and are generally also dielectric fluids, which behave differently toward polymers than do known oils used in automotive engineering. Also, some fluid components are volatile and go into the vapor phase if the lab test storage temperature is too high, while other fluids are hygroscopic. Fischer continues: “Complex, multi-layered materials, with possible interactions between components used in the immersion application, are also challenging.” These complex materials can include thermal interface materials (TIMs) (e.g., heat conduction pads or silicone and graphite mats); textile meshes (i.e., silicone sleeving); various adhesives; composite materials and metallic conductors such as copper conductor tracks; and aluminum or brass conductors with insulators of PVC, PE or high-performance coatings like polyamide-imide (PAI) or polyimide (PI). Fischer says, “Encapsulating compounds consisting of resins (i.e., epoxy, acrylic) also play an important role in these applications.” Additionally, the finished part geometries of the materials to be tested (e.g., cables, connectors) make determination of the corresponding mechanical properties difficult. 

Patrick Bauer is the expert team leader for PD thermal management fluids for Castrol Germany GmbH in Hamburg, Germany. Bauer notes, “As both immersion cooled batteries and immersion cooling in data centers are relatively new technologies, there is a lack of standardization to some extent in contrast to well established applications such as lubricants for internal combustion engines (ICEs) or transmissions. It might happen that unrealistic test methods from other applications might be chosen, which are not suitable for the actual applications in scope.”
 
Fischer notes that a better understanding of the material compatibility of various information technology (IT) components with corresponding oil formulations is being attempted by the Open Compute Project (OCP)³ addressing the lack of standardized test procedures. Bauer agrees that the OCP Material Compatibility in Immersion Cooling working group can be “a huge help to prevent or overcome such situations [by] aiming to define realistic and effective standards for the industry.” While new oil formulations, adapted temperatures, storage times and material parameters are being defined, even in the OCP, the definitions of these parameters may not be adequately narrow (e.g., material types defined for reference materials). 



Technical issues
Fischer identifies a number of issues related to the new immersion cooling oils and materials as needing further research and refinement:
• Are well-known ISO or ASTM lab testing standards still sufficient to date for evaluation of new immersion cooling fluids?
• Which test conditions are suitable on a laboratory scale (i.e., immersion time, temperature for accelerated aging, etc.)?
• How can defined standard materials be identified (i.e., suitable reference materials with known compositions)?
• How reproducible are the static laboratory tests?
• How accurate is the scalability of the laboratory results to the subsequent application?

Kevin Wirtz, a science advisor and global lead in industrial fluid systems at Cargill, Inc. in Minneapolis, Minn., notes: “All fluids have some limitations on material compatibility.” He says, “The most robust industries will utilize a selection of materials that are compatible with all families of cooling liquids [as] this strategy secures the supply chain and provides the highest degree of safety and equipment longevity as the industry changes.” 

Wirtz notes: “Biobased fluids help reduce the embodied carbon in immersion cooling applications both in computing and electric vehicle (EV) applications.” He observes: “While natural esters are not suited to EV applications where overnight standing temperatures reach below -10°C, biobased synthetic esters offer a larger reduction than obtained with petroleum-based products with a larger degree of circularity.” However, the biobased content in these fluids also poses a challenge for material compatibility, especially in direct immersion cooling, as these fluids tend to oxidize and polymerize.

When asked what discrepancies he sees with the different types of reference polymers for lab testing, Wirtz says, “The non-uniformity of gaskets and plastic compositions presents the highest procurement risks to manufacturers.” He provides the example of “PVC components or gasket polymers that can be made with different molecular weight ranges, differing polydispersities and different additives,” and notes that all of “these differences can lead to cracking, swelling or embrittlement depending on the immersion liquid.” He finds: “The risk is highest with printable plastic components as these tend to be more heavily formulated.” 

Testing and development
As previously mentioned, identifying polymer reference materials for compatibility testing to develop novel cooling oils is a challenge. Wirtz says, “While the development of cooling liquids utilizes samples of virgin polymers for screening compatibility, ultimately every component that is used in a system must be tested before the systems become commercial to verify that the differences between manufacturer formulations do not result in component degradation or interfering contamination of the immersion or cooling liquids.”

Accelerated aging methods can be used to test the compatibility of materials with novel developed dielectric fluids. Wirtz says, “Multiple industry consortiums have developed standardized conditions for evaluating immersion compatibility.” However, he says, “Test protocols may require more rigorous test conditions to match the conditions of the application.” 

Basil Bodemann is responsible for thermal fluids development at Shell Global Solutions (Deutschland) GmbH/Shell Technology Centre in Hamburg, Germany. Bodemann agrees that identifying polymer materials for compatibility testing “can be a bit tricky.” He observes: “On the one hand we always prefer to test customer materials directly; however, on the other hand the material samples available for doing so may not fit a norm.” 

Bodemann notes “conflict between the vast numbers of polymer materials available (both brands and composition) and the relatively high costs associated with testing each” material. He says this supports the idea for testing reference cases. Based on the reference case results, Bodemann says, “Further assessment on suitability can be associated with individual pairings.” He also notes that artificial intelligence (AI) is already gaining ground on covering these associations. 

Wirtz says, “Accelerated aging is useful for predicting long term effects and component life, but often the softening points of polymers limit the temperature of the aging conditions, which may require a deviation from standardized aging tests.” Wirtz notes: “The decision tree for selecting components requires a cross-functional knowledge of test limits, liquid and polymer chemistries, structural engineering and application conditions to provide the highest degree of safety across the lifetime of the system.” 

Laboratory testing
Bauer notes: “Standard testing around material compatibility involves immersion tests with standardized test specimens with materials such as elastomers, plastics as well as metals in scope for the application.” The main materials that are in scope include nitrile butadiene rubber (NBR), hydrogenated nitrile butadiene rubber (HNBR) and fluorine caoutchouc material (i.e., FKM, which is also known as fluorocarbon-based fluor elastomer material and is commonly known as fluoro rubber or fluorine rubber), different polyamide (PA, i.e., nylon) grades as well as copper. 

Bauer notes: “Testing typically starts with durations between 100 to 500 hours and might later be tested for longer durations.” Testing temperatures, depending on the application, can range from -40℃ to higher than 150℃. If required, Bauer notes that these tests can also be performed in the presence of water. Tests are also being performed with adhesives, glues and sealing agents. 

Bauer notes that an important aspect of thermal management fluids “is that the lifetime of the fluids in the application is significantly longer than” for other fluids. In data centers, he notes: “This can easily mean five to 10 years of fluid lifetime.” He observes: “Ensuring the compatibility with all wetted materials over such long-time spans is definitely challenging and takes a collaborative approach between the fluid manufacturer and the original equipment manufacturers (OEMs) and other suppliers along the value chain to ensure.” 

Bodemann says, “Static laboratory tests are a welcomed way for first order material/fluid selection” where having proven standards supports acceptability. He notes several questions that must be considered:
• How effective and reproducible are the static immersion pre-tests on lab scale? 
• How accurate is the transferability and scalability of these tests to the later application? 

Bodemann says, “For fluid suppliers pre-tests on lab scale are effective and serve as a decision basis.” He does observe that sometimes further discussions are needed to assess the genuineness of the results for the total system.

Selecting the right seals and adhesives
Bauer says, “The ideal way to correctly select seals and adhesives is to work collaboratively and as a part of co-engineering programs.” He states, “The question of which materials are to be used and to come into contact with our fluids is one of the first we bring up in project kick-off discussions.” Arriving at a mutual understanding regarding the test conditions is an important factor, “so that it can be ensured they are realistic and that materials are not considered further just because the test conditions were chosen poorly.” 

The impact of the tested materials on the fluids is another important factor to consider. Bauer notes: “This is why we typically also test the used fluid from those material compatibility tests” to assess that the material is compatible with the fluid. Knowing the materials that have been used in the past in indirectly cooled systems that are not suitable for use in immersion cooling is especially important for new applications such as immersion cooling in both automotive batteries and data centers. Bauer finds it especially important for battery and data center applications that the electrical properties of the fluids remain in specification when testing material compatibility of the materials with the fluids. 

Bauer observes that the challenge in interpreting results from seal compatibility testing of thermal management oils in comparison to currently well-known lubricants like engine and transmission oils lies in “the absence of a database of compatibility data.” This lack of data makes it challenging to identify the best oil and shows the need for increased testing of possible thermal management oils as well as materials. 

These are applications, Bauer observes, “where the electrical properties of the fluid are important and where it is important these are maintained even when interacting with a number of different materials.” Bauer identifies water as an example where “water that can agglomerate in or diffuse through plastics and find its way into the fluid” can be a problem. Also, thermal management fluids have increased the likelihood of an interaction between the fluid and materials, which therefore needs to be investigated.”

Batteries and data centers
Bodemann notes that the interface between fluid application in the cooling of EV batteries and data centers varies by application (i.e., direct or indirect cooling) and by fluid requirements. EV batteries and data centers both share that fluid is an effective heat transfer medium either for entire electric (sub-)systems or for concentration points. Bodemann says, “Lab compatibility tests remain crucial as these cannot be simulated computationally by any reasonable means.”

Further research questions
Bodemann says, “Material compatibility is one of the main challenges in technology adoption for advanced fluid cooling.” He notes: “Physical design parameters and the quest of heat transfer, even highly non-linear in nature, can be solved by aid of powerful computing; however, life cycle assessment on material suitability still relies on empiric testing.” The specific requirements that must be met are very much industry dependent as cultures can vary widely between industries.

Conclusions
This topic of immersion cooling overlaps with novel cooling fluids and oil formulations for electric cars (i.e., direct battery and e-motor cooling). Oil manufacturers are therefore also trying to establish themselves in other strategic business areas like data center cooling where oil formulations can be similar. Data center liquid immersion cooling is an increasingly relevant topic for all lubricant manufacturers that are also active in the thermal management of EVs. 

REFERENCES
1. ISO 1817:2024, (2024), “Rubber, vulcanized or thermoplastic — Determination of the effect of liquids,” Edition 8. Available at www.iso.org/standard/86602.html#lifecycle.
2. ASTM D471, ASTM D471-16a, (2021), “Standard test method for rubber property—effect of liquids.” Last Updated: June 25, 2021. Available at https://store.astm.org/d0471-16ar21.html. 
3. www.opencompute.org/

Andrea R. Aikin is a freelance science writer and editor based in the Denver area. You can contact her at pivoaiki@sprynet.com.