The volumetric thermal expansion coefficient β describes how much a fluid expands per degree of temperature rise at constant pressure:
β = −(1/ρ)(dρ/dT)
β is a derivative, and differentiating digitized data amplifies its noise. Each grade's density points lie on a straight line to R² ≥ 0.9998, with a worst-case departure of 0.78 kg/m³ — 0.075% of ρ. But consecutive anchor points sit only 45–75 °F apart, so that 0.05% wiggle becomes a 3.6–12.2% swing in the slope measured between one pair of points and the next. Taking β from those local slopes produced a sawtooth that rose, fell, and rose again — something a fluid whose density decreases monotonically cannot physically do.
So β here comes from a single least-squares straight line fitted to each grade's digitized points, the same approach the MIL-PRF-23699 page uses. With ρ = b + mT, that gives β = −m/(b + mT): a smooth curve that rises gently with temperature, because the same absolute density loss per degree is a progressively larger fraction of a shrinking density. That gentle rise is real; the steps were not.
For a fixed mass of fluid, volume is inversely proportional to density. The percent volume change relative to 77 °F (25 °C, Eastman's own reference temperature for density) is:
ΔV/V₀ = (ρ₀ − ρ(T)) / ρ(T) × 100%
Negative values indicate that the fluid is denser than at 77 °F (volume contracted). This is useful for sizing hydraulic reservoirs and accumulators across the full service temperature range.
β and ΔV/V₀ are derived entirely from the density curves already digitized on the Skydrol Density page — real measured curves for LD-4, 500B-4, and the obsolete Skydrol 7000 (the last extrapolated below its 7.65 °F digitized floor, shown dashed).
Skydrol PE-5 and Skydrol 5 are deliberately absent. Each has only a single published density value, at 25 °C, and β is a slope — one point cannot produce one. Substituting another grade's curve would mean publishing that grade's β under a different name, so this page shows only grades whose β is actually measured. The Prandtl and thermal diffusivity pages do estimate density for those two grades, because there density is one input among several rather than the answer itself; both state the assumption explicitly.