Three idealized paths bracket every real compression.
Isothermal compression, with perfect cooling, needs the least work. No machine achieves it, but it is the yardstick for intercooling:
Wiso = (Z R T1/M) ln r(3)
Isentropic (reversible adiabatic) compression is the reference used for reciprocating and screw machines:
Hs = Zavg (R T1/M) · k/(k − 1) · [r(k−1)/k − 1](4)
Polytropic compression follows P vn = constant with a fixed small-step efficiency ηp. It is the reference for centrifugal and axial machines:
Hp = Zavg (R T1/M) · n/(n − 1) · [r(n−1)/n − 1](5)
(n − 1)/n = (k − 1) / (k ηp)(6)
The discharge temperature follows from either definition:
T2 = T1 r(n−1)/n or T2 = T1 [1 + (r(k−1)/k − 1)/ηs](7)
and both routes give the same gas power and shaft power:
Pgas = ṁ Hp/ηp = ṁ Hs/ηs Pshaft = Pgas + mechanical losses(8)
The two efficiencies describe the same machine, but they are not equal. For r > 1 the isentropic efficiency is always the lower one, and the gap grows with ratio:
ηs = [r(k−1)/k − 1] / [r(k−1)/(k ηp) − 1](9)
Example 2 shows 80 % polytropic is 78.9 % isentropic at a ratio of 1.58. At a ratio of 8, the same 80 % polytropic is only about 74–75 % isentropic. Quote efficiencies without saying which kind and you are a few percent wrong on power before you start.
Why do centrifugal people prefer polytropic? Because the polytropic efficiency of a well-designed impeller hardly depends on its ratio, so stages can be stacked and compared. Isentropic efficiency falls as stages are added, even if every stage is equally good.
Real gases
Equations (4) and (5) assume an ideal gas corrected with an average Z. With Zavg = (Z1 + Z2)/2 and k at average conditions, they stay within about 1 % of a rigorous calculation for most hydrocarbon and air duties; Example 9 compares them for CO₂ up to 30 bara and finds 0.5 % on head. Near the critical point, in the dense phase or at very high ratios, use an equation of state (Peng–Robinson, BWRS or GERG-2008) in your simulator.
ASME PTC 10 and ISO 5389 define polytropic work using the methods John Schultz published in 1962, with compressibility functions X and Y.8 Those methods lose accuracy near the critical point and in the dense phase.9 The 2022 revision of PTC 10, its first since 1997, moved to real-gas relations, a polytropic process throughout, and new methods for polytropic head and efficiency.10 If a vendor's head disagrees with yours by more than 1–2 %, ask which method and equation of state they used before you argue about efficiency.