PG25 Is Not Water Plus 4.5% Volume

Substituting 25 percent propylene glycol for treated water changes the volumetric heat capacity from 4,146 to 3,968 kJ/(m³·K) at 40 °C, so the same heat at the same temperature difference needs 4.5 percent more flow [MODELED]. That figure is the smallest of the differences and the only one usually quoted. Carrying the same substitution through friction and convection gives pump power up 32 percent, pipe pressure drop up 26 percent and in-tube heat transfer coefficient down 31 percent [MODELED]. The 4.5 percent is not wrong. It is not the cost. The comparison holds at 40 °C on generic property correlations. No supplier property table for the fluid actually to be purchased has been obtained.

The 4.5% Flow Rise Is the Smallest of Four PG25 Ratios

Hold the heat, the temperature difference and the pipe diameter constant, and every difference follows from three fluid properties. Volumetric heat capacity sets the flow ratio at 1.045 [MODELED]. Kinematic viscosity, roughly 1.73 times that of water at 40 °C, drops the Reynolds number to 0.603 of the water case even though velocity has risen [MODELED]. Lower Reynolds number raises the Darcy friction factor to 1.135, and combined with density and the square of velocity the pipe pressure drop lands at 1.264 [MODELED]. Multiply pressure drop by flow and hydraulic pump power is 1.321 [MODELED].

Convection moves the other way and matters more. Conductivity, Reynolds number and Prandtl number together put the in-tube heat transfer coefficient at 0.685 of water — a 31 percent reduction on the fluid side of every plate and every cold plate. On a plate heat exchanger with water on the other side, that translates into an overall heat transfer coefficient 15 to 20 percent lower, which is to say the same exchanger moves 15 to 20 percent less heat at the same mean temperature difference [MODELED].

If Water-Rated, PG25 Cuts Exchanger Duty 15–20%

The exchanger result reaches the redundancy check directly. If a 1.5 MW CDU nameplate was established on water or treated water and the unit is filled with PG25, its effective capacity is roughly 1.20 to 1.28 MW [SENSITIVITY]. A 3+1 zone that was credited with 4,500 kW after losing one unit then has 3,600 to 3,825 kW, against a modelled zone load of 4,194 kW at a 0.90 liquid-cooling fraction [SENSITIVITY]. A margin of 7.3 percent becomes a shortfall of 9 to 14 percent, and the single-unit failure case stops closing. Nothing in the project record states which fluid the nameplate was calibrated on [HOLD].

First Energisation Runs Near the Laminar Boundary

A second effect appears at part load rather than at full load, which is why it tends to surface during first energisation. Viscosity governs whether rack branch flow stays turbulent. At 40 °C a branch holds Reynolds number above 20,000 down to about 33 percent of rated flow, reaches 10,000 at about 17 percent and leaves the turbulent regime near 7 percent [MODELED]. On a cold fill at 10 °C the same branch is already at Reynolds number 10,000 at 42 percent of rated flow [MODELED]. A facility commissioning its first racks at 10 to 20 percent IT load therefore runs closer to the laminar boundary than the same facility at full load [SENSITIVITY]. Throttling valves to hold a temperature difference makes it worse rather than better.

The controlled supplier-package review records the same viscosity problem from the other end of the temperature range: an ambient viscosity envelope does not cover a cold-start pump. It also records that concentration, the as-supplied formulation, and compatibility across a wetted set that spans more than one alloy family and at least one brazing filler remain outstanding. Material differences are not asserted as non-compliances. They are unclosed.

Limits and open items

Modelled: all four ratios and the exchanger derating band, computed at 40 °C from generic glycol-solution property correlations using standard friction and convection relations. They are internally consistent and independently reproducible from the property values. They are not supplier data. [MODELED]

Sensitivity, not a design value: the 1.20 to 1.28 MW effective capacity band and the resulting zone shortfall are conditional on the nameplate having been calibrated on water. [SENSITIVITY]

Open: the concentration basis, inhibitor package and as-supplied formulation; the temperature-dependent property table for the fluid actually purchased, including behaviour at cold-start rather than operating temperature; the fluid the CDU nameplate was calibrated on; and material compatibility across the wetted set. [HOLD]

No Engineer of Record has reviewed these ratios. No fluid is specified. No unit capacity is accepted. Nothing here releases equipment for purchase.

Ask for the nameplate fluid

The 4.5 percent figure is not wrong; it is the only one of the four that is small, and it is the one that gets written into the basis of design. A useful discipline is to require any glycol statement in a document to carry its companion: if a submission says the flow increases 4.5 percent and does not say what happens to the heat transfer coefficient, the submission has answered the easy third of the question. The one question worth asking the cooling supplier before anything else is which fluid, at which concentration and which temperature, sits behind the capacity on the nameplate.

© 2026 K&K Data Service Inc. All rights reserved. Reproduction or republication is permitted only with clear attribution to K&K Data Service Inc. and a working hyperlink to the canonical URL of this article. Excerpts must preserve the technical context, maturity labels, assumptions, and limitations. No excerpt may imply project approval, field validation, certification, or endorsement that the original article does not state.

Source: K&K Data Service Inc., “PG25 Is Not Water Plus 4.5% Volume,” https://www.kkdatasvc.com/lab/direct-liquid-cooling-and-cdu/pg25-is-not-water-plus-4-5-percent-volume/.