Immersion cooling fluids: what to confirm beyond breakdown voltage

The number on the first page of the datasheet

When an immersion cooling tank is being specified, the first property that gets compared is usually the breakdown voltage. It is easy to understand — a higher number means better insulation — and it appears prominently on every dielectric fluid datasheet.

The problem is that a breakdown voltage figure, taken in isolation, tells you almost nothing about whether the fluid will work in a real tank over a real service life. It is one electrical property measured under one set of laboratory conditions. It does not tell you whether the fluid will circulate at the tank's cold-start temperature. It does not tell you whether it will remain chemically stable after two years of contact with server materials at operating temperature. And it does not tell you whether it meets the fire-safety requirements of the facility in which the tank sits.

This article sets out the properties that should be on the same page as the breakdown voltage, using the two fluid families the Glacier catalogue carries for electronics cooling as reference points. It does not claim either family is approved by any equipment vendor, and it does not recommend a fluid for a specific tank design.

Three properties that belong in the same comparison

The LM-17 series is described in the catalogue as modified hydrocarbons, developed for server immersion cooling. The catalogue states the fluids have high breakdown voltage, high DC resistivity, good thermal and oxidative stability, resistance to yellowing and no corrosive sulfur. Table 15 gives three models:

If the comparison is made on breakdown voltage alone, LM-17C is the obvious choice. If viscosity at the tank's circulating temperature is included, the picture changes.

The three models differ in viscosity by a factor of more than six — 21.13 cP for the base model against 3.42 cP for LM-17C — and that difference directly affects pumping power, flow distribution within the tank and the uniformity of server temperatures. The model with the highest breakdown voltage also has the lowest flash point, at 107 °C, which determines whether the fluid satisfies the fire-safety requirements of the facility. And the model with the highest breakdown voltage has the warmest pour point, at −40 °C, which limits cold-start margin.

The LM-14 series takes a different chemistry route altogether, described as modified perfluorinated compounds. Table 12 lists six models spanning temperature ranges from −110 °C to +135 °C, with viscosities from 1.03 cP to 6.32 cP at 20 °C. The catalogue describes these fluids for electrical insulation, chemical inertness and precision temperature control, with applications including electronics and semiconductor equipment. The comparison between LM-14 and LM-17 is therefore not a comparison of two immersion fluids; it is a comparison of two different approaches to electronics cooling, and the architecture — indirect precision loop versus direct immersion tank — decides which family applies.

Pour point is not an operating limit

Catalogue Table 15 reports pour points, not operating-temperature minima. A pour point of ≤ −50 °C for LM-17A means the fluid will flow at that temperature in a laboratory test. It does not mean the fluid will pump efficiently through a distribution manifold at −50 °C, or that the tank can be started from cold at that temperature without a pre-heat cycle.

The industry practice is to apply a margin — typically 10 °C to 15 °C above the pour point — as the practical minimum circulating temperature. Below that margin, viscosity rises to a point where pump power, flow distribution and heat transfer become unacceptable before the fluid actually stops moving. The catalogue's viscosity figures, given at 20 °C, are a reference point for selection; they do not describe the fluid's behaviour at the cold end of the tank's operating envelope without additional data.

Single-phase versus two-phase — a fluid designed for one is not designed for the other

The distinction matters because the physical demands differ. A single-phase immersion fluid circulates as a liquid, removing heat by sensible temperature rise. A two-phase fluid boils at the heat source, and the condenser is designed around that boiling point.

A fluid designed for single-phase immersion cannot be assumed suitable for a two-phase tank, and vice versa. The catalogue describes LM-17 as a single-phase immersion fluid with a closed-system usage instruction: clean and blow-dry the system before filling, add the product directly, and keep the system closed during use to prevent moisture ingress. It does not describe it as a two-phase fluid, and it does not give a boiling-point specification that would support two-phase design.

The LM-14 series, by contrast, lists boiling points from 50 °C for LM-14A to 150 °C for LM-14G. Those are boiling points under the test conditions stated in the catalogue, and a two-phase system would need to be designed around them. The catalogue does not describe LM-14 as an immersion fluid, and the boiling-point data should not be taken as a two-phase design specification without confirming the test conditions, pressure and single-phase versus two-phase operating mode with the manufacturer.

What to confirm before you fill the tank

• The complete electrical specification, not just breakdown voltage. DC resistivity, dielectric constant and dissipation factor matter for the components that are submerged. Confirm the test methods and the conditions under which the values were measured.
• Viscosity at the tank's lowest circulating temperature, with a margin below that for cold-start conditions.
• The flash point against the facility's fire-safety requirements. Published guidance for single-phase immersion fluids typically requires a flash point above the facility's maximum credible temperature, and some facility standards set a minimum above 130 °C.
• Material compatibility with every wetted component. This includes server boards, solder, conformal coatings, connectors, cables, pump seals, tank materials, gaskets and any instrumentation in contact with the fluid. The catalogue's statement that LM-17 does not contain corrosive sulfur is one compatibility claim; the full list requires the manufacturer's written confirmation.
• Thermal and oxidative stability under the tank's operating conditions. The catalogue describes the LM-17 series as having good thermal and oxidative stability and resistance to yellowing. Confirm the expected service life and the monitoring programme for acid number, viscosity and appearance.
• The regulatory position of the fluid's chemistry. Fluorinated fluids are currently the subject of active regulatory proposals in multiple jurisdictions, and a fluid that is available today may be subject to reporting obligations or substitution requirements during the life of the tank. The PFAS regulation summary sets out the current state of the proposals; confirm the exact product's status before committing to a multi-year fill.
• The disposal path. Immersion fluids are a waste stream at end of life. Confirm the disposal route and cost with the supplier and the site's environmental permit.

What this article does not settle

It does not recommend a fluid, a tank architecture or an operating temperature. Those decisions are made by the server vendor, the tank designer and the facility operator, and they depend on the specific hardware and operating conditions.

It also does not claim that any Glacier fluid is approved by any equipment vendor, or that a breakdown voltage figure establishes dielectric suitability for a specific server design. Equipment approval, warranty and safety certification rest with the hardware vendor, not the fluid supplier. The catalogue provides reference properties; the decision to use a fluid in a specific tank requires the equipment vendor's written confirmation and the manufacturer's current technical data sheet and safety data sheet.

The electronics cooling solution page sets out how these fluid families map onto the different architectures, and the indirect liquid cooling versus immersion article covers how the architecture decides which fluid questions apply.