The carbon-capture advantage of semi-closed supercritical CO 2 (sCO 2 ) cycles originates from oxy-fuel combustion; however, the additional energy consumption of air separation units (ASU) significantly reduces net system efficiency, necessitating further efficiency improvement. This work proposes an innovative cold-end heat rejection methodology that employs mixed-flow direct-contact condensation between CO 2 and water coupled with density-difference separation, thereby replacing conventional gas–liquid indirect cooling with liquid–liquid cooling. This approach reduces the heat-transfer temperature difference, enables subcritical CO 2 liquefaction, and transitions the compression mechanism from compressor-driven near-critical gas-phase compression to pump-driven liquid-phase compression. The thermophysical feasibility rests on CO 2 –H 2 O mutual solubilities of 25 mol/kmol and 4 mol/kmol, respectively, coupled with a liquid-phase density differential of approximately 260 kg/m 3 across the operating envelope, which together facilitate the coupling of mixed-flow condensation and density-difference separation. A semi-closed sCO 2 cycle incorporating this concept (MSC) is constructed and evaluated. Results demonstrate a cold-end temperature reduction from 32 °C to 27 °C, a 67.35 MW decrease in compression power, a 4.7 percentage-point increase in cycle thermal efficiency to 65.51%, and an improvement in system exergy efficiency from 57.66% to 59.76%. This cold-end-focused study provides a novel technical pathway for performance enhancement of semi-closed CO 2 cycles. • Constructs a mixed-flow semi-closed sCO 2 cycle using direct-contact condensation and density-difference separation. • Achieves CO 2 liquefaction with small temperature-approach heat rejection, enabling pump-driven liquid pressurization. • Develops a CO 2 –H 2 O mutual-solubility correlation and validates phase splitting via multiphase CFD modeling. • Delivers a 4.72%-point thermal-efficiency gain (to 65.53%) and a 67.45 MW reduction in compression power.
Sun et al. (Wed,) studied this question.