Carnot Efficiency & Heat-Engine Calculator: η, W, Q, ΔS

Classical thermodynamics holds up beautifully across most of the engineering world, with a few well-defined cracks. The Carnot ceiling η = 1 − T_c/T_h is exact for any reversible heat engine, no matter the working fluid, no matter the cycle. That single inequality decides what coal plants, gas turbines, jet engines, refrigerators, and car engines can do, and it's why thermal generation hasn't crept past about 65% efficiency in any commercial unit. The 1824 Carnot result, restated by Clausius and Kelvin in the 1850s as the second law, hasn't been overturned in 200 years.

The breakdown points are specific. Phase change: at the boiling shelf, ideal-gas math gives wildly wrong answers because the latent heat dwarfs the sensible-heat capacity (water at 100°C and 1 atm: L_v = 2,257 kJ/kg). Real Rankine analyses use steam tables, not PV = nRT. Real-gas behavior at high density or low temperature: the compressibility factor Z = PV/(nRT) drifts away from 1, and you need cubic equations of state (Peng-Robinson, SRK) or NIST's reference data. Quantum and statistical regimes: at very low temperature, the third law of thermodynamics (Nernst) and quantum heat capacities take over from classical equipartition. Solid heat capacity at room temperature is the Dulong-Petit value 3R, but at low T it falls as T³ (Debye model). Specific heats stop being constants of nature and become functions of T.

This calculator solves the four problems that come up most often. Carnot efficiency directly. Real engine efficiency from measured Q_h and Q_c. Work and heat for an ideal gas through one of the four canonical processes (isothermal, adiabatic, isobaric, isochoric). And entropy change of an ideal gas across two states. The page isn't a substitute for steam tables (IAPWS-IF97) or REFPROP-grade real-gas calculations. It's a fast, correct sanity check for homework, exam prep, and the classroom version of cycle analysis.

One thing this tool doesn't do: chemical reaction spontaneity. If the problem is ΔG = ΔH − TΔS for a reaction, equilibrium constants, or van 't Hoff plots, that's a chemistry tool and lives at /tools/chemistry/gibbs-free-energy-equilibrium-calculator. The shared symbol ΔS doesn't mean it's the same problem. Reaction Gibbs energy asks "will this reaction go forward". Carnot efficiency asks "how much shaft work can I extract from this heat". Different question, different tool.

Notation throughout: T is absolute temperature in Kelvin, P is pressure in pascals or bars, V is volume in cubic meters or litres, n is moles, R = 8.314 J/(mol·K), C_v and C_p are molar heat capacities, and γ = C_p/C_v is the heat-capacity ratio (1.40 for air at room temperature, 1.67 for monatomic helium, 1.30 to 1.33 for steam in normal Rankine ranges).

The gravitational analog here is the convention used for the zero of energy. Thermo doesn't have gravitational potential per se, but it shares the same structural problem of choosing where zero lives. In gravity you set U(∞) = 0 so that bound orbits have negative total energy. In thermodynamics you face two analogous choices: where zero of internal energy U sits, and which direction counts as positive for work and heat. Both are conventions, and both matter for closed-cycle bookkeeping.

Zero of internal energy. For an ideal gas, U is a function of T only, and we usually take U = nC_v T relative to U(T = 0) = 0, just as Newtonian gravity takes U(∞) = 0. The choice doesn't affect any ΔU you actually compute, because ΔU is path-independent for a state function and the offset cancels. For real fluids and especially for phase-change problems, NIST and IAPWS pick reference states that suit their tables (saturated liquid at the triple point of water for IAPWS-IF97, ideal-gas zero at 0 K for many gas tables). If you mix tables, check the reference state. A 2,257 kJ/kg discrepancy in a steam-cycle analysis is almost always a reference-state mismatch, not a physics error.

Sign convention for work. Two camps exist and they disagree. Physics convention: W is work done by the gas, so W > 0 means the gas pushed the piston out and did work on the environment. The first law reads ΔU = Q − W. Engineering convention (used in most thermo textbooks like Cengel and Boles): W is work done on the gas, so W > 0 means the environment compressed the gas. The first law reads ΔU = Q + W. Both versions give the same physics. Pick one and stay with it. This calculator uses the physics convention (W done by the gas), which matches typical undergraduate physics textbooks.

Sign convention for heat. Q is positive when heat flows into the system, negative when heat leaves. This convention is universal across both physics and engineering. A gas absorbing heat from a hot reservoir has Q > 0; a gas dumping heat to a cold reservoir has Q < 0. Carnot bookkeeping splits Q into Q_h (positive, absorbed from the hot reservoir) and Q_c (positive, rejected to the cold reservoir, even though as a system-side flow it would be negative). Watch the labels.

Why the conventions matter for cycles. A closed cycle (the system returns to its starting state) has ΔU_cycle = 0. So the first law gives Q_net = W_net. The net heat absorbed equals the net work done. For a heat engine, W_net > 0 and Q_h > |Q_c|, so efficiency η = W_net/Q_h = 1 − Q_c/Q_h. For a refrigerator, you reverse: W_net < 0 (work goes in) and the COP for cooling is COP_c = Q_c/|W_net|. For a heat pump, COP_HP = Q_h/|W_net| = COP_c + 1. All four (η, COP_c, COP_HP, and reversibility) follow from the same convention choices applied consistently. Mix them up and you'll get η > 1, which always means a sign error rather than a violation of the second law.

Absolute Kelvin is forced. Carnot efficiency is η = 1 − T_c/T_h. That's a ratio of temperatures, and ratios only make sense on an absolute scale. A coal-fired boiler at 540°C rejecting to a 30°C condenser, computed in Celsius, gives η = 1 − 30/540 = 94%. In Kelvin: T_h = 813, T_c = 303, ratio 0.373, η = 62.7%. Off by 30 percentage points. The Kelvin scale isn't convenience; it's built into the definition of thermodynamic temperature via the Carnot ratio (Kelvin's 1848 derivation). Differences of T are scale-independent (so heat-conduction and thermal-expansion problems can use Celsius if you're careful), but anything with a ratio of T needs Kelvin.

The thermo analog of Kepler's diagrams is the P-V diagram for a closed cycle. Kepler's ellipses gave geometric shape, period, and orbital energy. A Carnot or Otto or Rankine cycle on a P-V plane has its own geometry (closed loop), its own period (the cycle time), and its own energy bookkeeping (area enclosed equals net work per cycle). Same problem class, different state space.

Cycle geometry. A closed cycle is any sequence of processes that returns the working fluid to its starting state. On a P-V diagram, the cycle is a closed curve. The area enclosed is the net work per cycle, |W_net|. If the curve runs clockwise (work output), it's a heat engine. If it runs counter-clockwise (work input), it's a refrigerator or heat pump. Four canonical cycles dominate undergraduate teaching:

• Carnot: two isotherms (at T_h and T_c) connected by two adiabats. The most efficient possible engine between any two reservoirs. η = 1 − T_c/T_h. Not used in real engines because the isotherms require infinite heat-transfer time at finite ΔT.
• Otto: two adiabatic strokes, two isochoric strokes. The idealization of a gasoline engine. η = 1 − 1/r^(γ−1), where r = V_max/V_min is the compression ratio.
• Diesel: two adiabats, one isobaric heat addition, one isochoric heat rejection. Differs from Otto in the heat-addition stroke. Slightly less efficient than Otto at the same compression ratio, but real Diesel engines tolerate higher r.
• Rankine: isobaric boiler, adiabatic turbine, isobaric condenser, adiabatic pump. The steam-power-plant cycle. Crosses phase boundaries, so it needs steam tables rather than ideal-gas math.

Cycle period (a.k.a. cycle frequency). Real engines run at finite speed, and the cycle period sets power. A four-stroke gasoline engine at 3,000 rpm completes a full cycle every two crankshaft revolutions, so 1,500 cycles per minute, or 25 Hz. The work per cycle times the cycle frequency gives mechanical power. A 2.0 L engine producing 250 J of net work per cylinder per cycle, with four cylinders at 25 Hz, delivers 4 × 250 × 25 = 25,000 W = 25 kW shaft power. Steam plants run continuously rather than in discrete cycles, but the analogy holds: mass flow rate times work per kg of steam gives shaft power. A 600 MW coal plant moves about 500 kg/s of steam through its cycle.

Energy per cycle. For a closed cycle, ΔU_cycle = 0, so Q_net = W_net. The net heat absorbed equals the net work done. Carnot, Otto, Diesel, and Rankine all obey this. What differs is how Q is split between Q_h (absorbed from a hot reservoir) and Q_c (rejected to a cold reservoir) and how that ratio relates to T_h and T_c. For Carnot, Q_c/Q_h = T_c/T_h exactly. For real cycles, Q_c/Q_h > T_c/T_h, which forces η < η_Carnot. The geometric picture is that any non-Carnot cycle has parts of its boundary at intermediate temperatures, and those segments contribute extra entropy that the second law charges to the engine's efficiency.

Closed cycle on a T-S diagram. The same cycle drawn on a temperature-entropy plane gives a different geometric story. Carnot is a rectangle (two horizontal isotherms, two vertical adiabats with ΔS = 0). The area enclosed equals the net heat (= net work). The maximum possible area for a given pair of T_h and T_c is the rectangle, which is why Carnot is the upper bound. Real cycles are squashed shapes whose area is always less than that rectangle for the same temperature limits. T-S diagrams are the natural tool for visualising why Otto, Diesel, and Rankine all underperform Carnot.

Run the orders of magnitude before deciding what level of model to use. Real cycles, real numbers.

The lesson: when you raise T_h or lower T_c (or both), Carnot rises and real engines track. That's why power plants try to dump waste heat into the coldest available reservoir (rivers, oceans, seawater intakes), and why thermal-plant output measurably drops during summer heat waves. A 4 K rise in cooling-water temperature cuts net efficiency by roughly half a percentage point, which over a year is a meaningful number of MWh.

Worked example: 1 mol N₂ heated from 300 K to 600 K at constant pressure. N₂ is diatomic, so γ = 7/5 and C_p = (7/2)R = 29.1 J/(mol·K), C_v = (5/2)R = 20.8 J/(mol·K). Going through an isobaric process at P = 1 bar:W = nR ΔT = (1)(8.314)(300) = 2,494 J (work done by the gas as it expands).Q = nC_p ΔT = (1)(29.1)(300) = 8,730 J (heat absorbed).ΔU = nC_v ΔT = (1)(20.8)(300) = 6,240 J. Check: Q = ΔU + W = 6,240 + 2,494 = 8,734 J. The 4 J difference is rounding (C_p = C_v + R, exact for an ideal gas; here 20.8 + 8.314 = 29.114 vs. tabulated 29.1).The gas does about 29% of its absorbed heat as work (W/Q = 2,494/8,730 = 0.286). That's the isobaric expansion work share.

The four canonical processes. An ideal gas going from state 1 to state 2 can take any path, but four are named: isothermal (T constant; W = nRT ln(V₂/V₁), ΔU = 0, Q = W), adiabatic (Q = 0; PV^γ = const, W = (P₁V₁ − P₂V₂)/(γ − 1), ΔU = −W), isobaric (P constant; W = PΔV, Q = nC_p ΔT, ΔU = nC_v ΔT), and isochoric (V constant; W = 0, Q = ΔU = nC_v ΔT). Mix these legs to build any closed cycle. The work numbers differ by path; the area under the PV curve gives W for each leg.

The framework labels for this slot were written for the gravitational tools in this cluster. The thermo analog is "the boundary cases where the second law and the ideal-gas approximation themselves break down or hit hard walls". Three matter.

The third law and absolute zero. Walther Nernst's third law of thermodynamics, formulated in 1906 and refined through the 1920s, says that the entropy of a perfect crystal at absolute zero is zero (or a universal constant) and that absolute zero cannot be reached in a finite number of steps. This puts a hard wall on cooling. Adiabatic demagnetisation, dilution refrigerators, and laser cooling can reach below 1 nK in the lab (Bose-Einstein condensates achieved 0.5 nK at MIT in 2003), but never zero. Carnot efficiency formally goes to 100% as T_c → 0, but the work required to maintain T_c near zero rises faster than the efficiency gain, so the limit is unreachable in practice. This is the thermodynamic version of the light-speed wall: the formula keeps giving sensible numbers, but the kinematics says you can't actually get there.

Black holes have entropy. The Bekenstein-Hawking result S_BH = (k_B c³ A)/(4 G ℏ) gives a black hole entropy proportional to its event-horizon area. For a solar-mass black hole, A ≈ 1.09×10⁸ m² and S_BH ≈ 1.5×10⁵⁴ J/K. That's vastly more entropy than the Sun has now (~10³⁴ J/K), and it implies that black holes radiate (Hawking radiation, 1974) at a temperature T_H = ℏc³/(8π G M k_B), which for a solar-mass hole is 6×10⁻⁸ K, far below the 2.7 K cosmic microwave background, so they net-absorb. Stellar black holes don't evaporate on any timescale shorter than the age of the universe. Primordial black holes of less than ~10¹² kg would be evaporating now, and that's an active observational target for high-energy gamma-ray telescopes. Classical thermodynamics has nothing useful to say about this regime; you need quantum gravity.

Bound vs. unbound: phase change and real-gas behavior. Ideal-gas thermodynamics works when the working fluid is dilute and weakly interacting. The compressibility factor Z = PV/(nRT) measures how far off you are. Z = 1 is ideal. For air at room conditions, Z = 1.0006, which is fine. At 100 bar and 200 K, Z drops to about 0.74 for N₂, and the ideal-gas calculation is wrong by 26%. Modern engineering uses cubic equations of state (Peng-Robinson, Soave-Redlich-Kwong), virial expansions, or NIST's reference data (REFPROP, GERG-2008 for natural gas). Phase change is the worst case: at the boiling shelf, T and P stay constant while V changes by a factor of a thousand or more (water at 100°C and 1 atm: 1.04 mL/g liquid to 1,673 mL/g vapor, ratio 1,610). The latent heat L_v = 2,257 kJ/kg dwarfs sensible-heat capacity, so any cycle crossing a phase boundary needs steam tables (IAPWS-IF97), not ideal-gas math. A Rankine analysis with this calculator's ideal-gas tools is at best a sanity check on the Carnot bound, not a real efficiency estimate.

Where this calculator stops being useful. Two flags. First, condensing fluids in heat exchangers (refrigerants in evaporators, steam in condensers) have constant-T phase-change sections that the ideal-gas treatment misses. The LMTD design method we cover at /tools/physics/heat-exchanger-lmtd-helper handles those sections separately. Second, reactive systems (combustion, fuel cells, batteries) need chemical Gibbs energy on top of physical thermodynamics, which is the chemistry tool at /tools/chemistry/gibbs-free-energy-equilibrium-calculator. If your problem stays clear of phase boundaries, density is moderate (Z within a few percent of 1), and design-grade accuracy isn't the goal, the ideal-gas treatment here is fine. Otherwise, switch to property tables.

A 2.0 L naturally aspirated gasoline engine with compression ratio r = 10 is the workhorse of the family-car market. Take peak combustion temperature T_h = 2,500 K (close to adiabatic flame temperature for stoichiometric gasoline-air, slightly cooler in practice), exhaust gas T_c = 850 K (typical for naturally aspirated engines), and γ = 1.4 for air-standard analysis. Three efficiency numbers tell the story.

Comparison: Rankine cycle for a coal plant. Steam at 540°C (813 K), condensing at 30°C (303 K). Carnot bound: η = 1 − 303/813 = 62.7%. Real subcritical plants run 33 to 38%, supercritical plants reach 38 to 43%, ultra-supercritical pushes 45%. Why so much worse than Carnot? The Rankine cycle isn't Carnot: the boiler adds heat over a range of temperatures (preheat, evaporation, superheat) rather than at a single T_h, and the average effective T_h is much lower than the peak. Plus: turbine isentropic efficiency 88 to 92%, pump losses, generator losses, parasitic auxiliary loads. The current efficiency record for commercial generation is the combined-cycle gas turbine (Brayton topping a Rankine bottoming cycle), with Siemens claiming 63.08% in 2016 with the SGT5-8000H, and the bar has crept up since.

Comparison: residential refrigerator. T_c = 4°C (277 K) inside, T_h = 25°C (298 K) kitchen. Carnot COP_c = T_c/(T_h − T_c) = 277/21 = 13.2. Real residential fridges run COP = 2 to 4, because the compressor isn't adiabatic, the evaporator and condenser run at finite ΔT, refrigerant lines have pressure drop, and the working fluid (R-134a, R-600a, R-290) has its own thermodynamic non-idealities. A heat pump for home heating from 0°C outdoor to 20°C indoor has Carnot COP_HP = T_h/(T_h − T_c) = 293/20 = 14.65 and real COPs of 3 to 5. Ductless mini-split heat pumps under mild conditions are the consumer products that come closest to Carnot, which is exactly why they're the cheapest electric heat per delivered BTU.

The pattern across all three examples is consistent. Carnot tells you the bound. Air-standard or ideal-cycle analysis tells you the bound for that specific cycle architecture. Real-engine performance is well below both, and most of the gap is real-engine losses, not cycle-architecture losses. This is why automotive engineering chases pumping losses, friction, and combustion completeness rather than chasing higher peak temperatures: you've already left more on the table from the engineering side than the thermodynamic limit can give back.