Joule Heist search

topics · updated 2026-09-25

Solar Cooling: PV vs Solar-Thermal — Is There a Better Option?

confidence: high volatility: warm verified: 2026-09-25fresh

Answers 'is there a better option for solar thermal cooling?' For most applications, yes: PV + electric heat-pump/vapor-compression now out-competes solar-thermal sorption cooling, because a vapor-compression COP of ~3-8 stacked on cheap PV converts far more roof sunlight into cold than a sorption chiller capped near COP 0.7-0.8, while being simpler and faster to pay back — IEA SHC Tasks 53 and 65 both attribute the shift to the PV price collapse. Solar-thermal cooling keeps a real niche where free or high-grade heat already exists, demand is steady, or cheap thermal storage beats batteries; a measured same-machine test shows the sunlight-to-heat path is ~4x more efficient than sunlight-to-electrons into a sorption cooler, which is exactly why hybrids (PV/T) and better collectors/cycles are the strongest within-thermal upgrades.

Short answer: for most applications, yes — PV-driven cooling has become the better option. The decade-long collapse in PV module prices flipped the economics: solar PV driving an electric heat pump / vapor-compression chiller now converts more sunlight into cooling, more cheaply and simply, than solar-thermal collectors driving a sorption (adsorption/absorption) chiller. Solar-thermal cooling is no longer the default — but it still wins in a specific, identifiable niche.

This article frames the choice the rest of the solar-thermal and solar-adsorption threads assume, and weighs the alternatives.

The core physics: why PV pulls ahead

Two architectures turn sunlight into cold:

A — Solar thermal + sorptionB — PV + vapor-compression
Pathcollector heats water → desorbs sorbent → sorption chillerPV makes electricity → electric chiller/heat pump
Cooling COPadsorption (solid sorbent) well below 1; absorption (liquid, LiBr or ammonia-water) single-effect ~0.7-0.8, double-effect up to ~1.35 with concentrating collectorsvapor-compression ~3-8 (Caferra’s PV chiller: ~3)
Solar capturecollector-to-generator ~35% measured (Selvaraj), higher for good collectorsPV module ~20%; ~9% measured end-to-end through a battery (Selvaraj) — but it feeds a high-COP machine
Balance of systemhot + chilled loops, large heat rejection, cooling-tower water, parasitics, backup heatone electrical connection, smaller heat rejection

Two thermal cycles, not one. “Sorption” on the thermal side is really two distinct cycles, and they are not interchangeable:

  • Adsorption — a solid sorbent (silica gel, zeolite, activated carbon) adsorbs refrigerant vapor onto its surface. Intermittent/batch, low COP, but it runs on low-grade heat — the data-center case below runs a silica-gel/water chiller on rack coolant at 40-65 °C. This is the low-temperature-waste-heat cycle (details in Adsorption Cooling).
  • Absorption — a liquid sorbent (LiBr/water or water/ammonia) absorbs refrigerant into the bulk liquid. Continuous, single-effect COP ~0.75 (Caferra’s Fresnel-driven unit), and the only route to double-effect operation — Task 65’s Fresnel-driven SunBeltChiller reaches an overall COP of 1.35 with storage, which needs a concentrating collector and a higher drive temperature.

So “double-effect” is an absorption-only capability; adsorption does not reach it. Keep them separate — most studies below are specifically one or the other (the data-center waste-heat case is adsorption; the sunbelt, Caferra, Madinah, and Selvaraj cases are absorption).

The decisive term is the ~4x-or-more COP gap. Even though PV captures a smaller fraction of incident energy than a thermal collector, multiplying ~20% PV efficiency by a vapor-compression COP of 3-8 delivers more cold per roof area than ~35-50% collector efficiency times a single-effect sorption COP of ~0.7-0.8 (and it still beats double-effect absorption’s ~1.35). PV cooling is also mechanically simpler: no second hydraulic loop, far smaller heat rejection, less water and parasitic power, and no gas-boiler backup to hold generator temperature.

What the authoritative reviews conclude

  • IEA SHC Task 53 (2018 position paper) named the cause outright: solar-thermal cooling’s market struggles are attributed to falling solar PV prices, which allowed cost-effective operation of vapor-compression chillers powered by grid-connected solar inverters; new market dynamism is “mainly boosted by PV driven systems,” and large Asian makers (Midea, Gree, Panasonic, LG) already integrate PV into heat pumps. The best solar-cooling systems of either kind reach an electrical SPF of 20-40 (>80% AC electricity cut).
  • IEA SHC Task 65 (2024, the Sunbelt field program: 32 projects, 18 countries, ~17 MW cooling) is blunter: PV-supported cooling “has become the dominant type of solar cooling system globally due to its simplicity in installation and low cost,” and “can result in lower levelized cost of cooling compared to solar thermal” (load- and weather-dependent). Only ~2,000 solar-thermal cooling systems exist worldwide, mostly customized, vs “several millions” of PV-driven units in Australia alone. Among the deployed thermal demonstrations, evacuated-tube (~30%), flat-plate (~17%), Fresnel (~17%) and parabolic-trough (~10%) collectors dominate; PV-driven systems were ~10% of the demos.
  • A 2025 head-to-head (Caferra et al., Energies) of a Fresnel-driven single-effect absorption chiller (COP ~0.75) vs a PV compression chiller (COP ~3), both ~500 kW: the ~4x COP gap makes PV-compression win for most demand profiles, especially variable/intermittent loads with grid export; absorption is competitive only with existing/waste heat or steady high summer load, and its gas-boiler backup hurts both emissions and economics.
  • A 2026 measured same-machine test (Selvaraj et al., Scientific Reports) drove one 60 W ammonia-water sorption refrigerator alternately from a flat-plate collector and from a 100 W PV module + battery under Indian tropical conditions. Thermal drive: evaporator 12 °C, COP 0.14, 35.3% collector-to-generator efficiency, generator peak 88 °C, stable output from thermal inertia. PV drive: evaporator 9 °C (colder), COP 0.12, ~9% end-to-end efficiency, cloud-sensitive swings. The sunlight-to-heat path is ~4x more efficient into a sorption machine — and that coexists with PV + vapor compression winning the system economics, because the electrons go to a COP-3+ machine instead. (Caveat: absorption, not adsorption; transfer qualitatively. The 88-89 °C generator temperatures are squarely in the flat-plate band the KB’s adsorption articles assume.)

Economics: capex and payback decide it

Both routes have near-zero fuel cost, so upfront investment dominates the levelized cost of cooling (LCOC):

  • Solar-cooling capex is still 1.5-2.5x conventional AC, with amortization of 8-15 yr (≈5 yr in favorable cases); a ~30% capex cut would put solar-driven LCOE always below the reference (Task 53). PV systems are “predestined” to scale cheaply alongside mass-market chillers and heat pumps, which is where the cost gap opens.
  • Task 65’s life-cycle analysis (heat at 150 EUR/MWh, gas at 45 EUR/MWh) finds hybrid solar-thermal configurations paying back in <~7 yr, with carbon pricing materially improving thermal’s competitiveness — i.e. thermal is not uneconomic, just usually beaten.
  • But solar-thermal can be very cheap in the right place: a 2025 Scientific Reports study of an evacuated-tube-assisted absorption chiller in Madinah, Saudi Arabia (32% solar fraction, CAPEX $90,223, OPEX ~1%/yr) reported LCOC $0.028/kWh, simple payback 3.04 yr, NPV $261,967 over 25 yr, IRR 32.2% — high irradiance + high ambient + a large steady load is where thermal economics shine. The authors’ own framing: a favorable-case site, not a general verdict.

Where solar-thermal cooling still wins

The thermal route remains the best option under a clear set of conditions — the decision is “free heat + steady load + cheap storage,” not raw efficiency:

  1. Free or high-grade waste/process heat already on site. A silica-gel/water adsorption chiller on a data-center water-cooled rack loop paid back in ~285 days for a 90 kW unit (379 days at 30 kW; Gupta & Puri 2021), with 22.5% better energy-savings efficiency and ~104 t/yr CO₂ avoided — unbeatable when the heat is free. Raising rack inlet from 40 to 65 °C improved chiller efficiency 16% at a 6% compute-per-watt cost, quantifying the thermal sweet spot.
  2. Trigeneration / CHP where engine or industrial exhaust would otherwise be dumped — the “existing heat source” case in which Caferra finds absorption competitive.
  3. Large, steady, high-load duty (industrial process cooling at high solar fraction — Task 65’s strongest thermal case; two-stage/medium-temperature absorption gives the best efficiency-cost balance for the Sunbelt).
  4. Cheap thermal storage beats batteries, or no monetizable PV grid export exists — Caferra’s general pattern is PV-compression for variable demand plus export, absorption plus thermal storage for continuous loads.
  5. Co-supplying low-temperature heat (domestic hot water) year-round alongside cooling, in moderate climates with easy thermal storage (Task 53’s named niche).

The “third options” — hybrids and better collectors

The most interesting answers aren’t “A or B” but hybridize:

  • PV/T (photovoltaic-thermal) collectors co-generate electricity and heat from one aperture, cooling the cells (raising PV yield) with the fluid that drives a thermal cycle. The 2023 Energies review of PV/T-driven cooling reports: absorption on PV/T max COP 0.615 (min payback 9.3 yr); adsorption on PV/T up to 81.7 kWh/day of electricity and an effective combined COP up to 2.1 (min payback 8.45 yr); and PV/T + desiccant + vapor-compression at overall COP 0.68 — beating both evacuated-tube+absorption (0.34) and plain PV+vapor-compression (0.29) on the review’s like-for-like basis. Absorption is the most mature and commercially adaptable of the four PV/T-driven families (absorption, adsorption, ejector, desiccant).
  • Better collectors + higher-effect cycles lift thermal performance (these are absorption cycles — adsorption has no equivalent): evacuated-tube and CPC collectors serve single-effect units (the Madinah case), while Fresnel/concentrating collectors enable modified double-effect machines like the SunBeltChiller — COP 0.35 at high daytime ambient, ~0.75 at night with dry recooling, overall up to 1.35 with storage — addressing single-effect chillers’ struggle in hot climates.
  • Desiccant and passive add-ons — the PV/T review’s desiccant + vapor-compression hybrid (see Desiccant & Free Cooling) splits the latent load off the chiller; radiative sky cooling is the KB’s other passive lever, cross-referenced here rather than sourced in this article.

Verdict

There is no universal winner, but the center of gravity has moved:

  • Default / most buildings: PV + heat pump is the better option — cheaper, simpler, faster payback, higher cooling-per-area, and the globally dominant solar-cooling type.
  • Low-grade free waste heat (~40-65 °C loops): adsorption still wins — the data-center / waste-heat case this KB centers on remains valid where the heat is genuinely free and continuous.
  • Higher-grade heat, steady large load, or sunbelt: absorption is the thermal route — higher COP, and the only one reaching double-effect; best for industrial/district cooling and hot climates.
  • Best frontier bet: PV/T hybrids and better collectors/higher-effect cycles, which capture both energy streams instead of choosing. The measured ~4x path-efficiency advantage of heat over electrons into a sorption machine is exactly why co-generating both from one aperture is attractive.

For a project like this KB’s waste-heat-driven strategy, the lesson is precise: heat-driven cooling is justified by the free waste heat, not by beating PV on a bare rooftop. On a bare rooftop with no waste heat, PV+heat-pump is usually the better option.

Provenance note. Earlier drafts cited a Hefei capex comparison, “older thermo-economic” payback figures, and a 2026 RSER off-grid review; none of those are in this article’s ingested sources and they have been dropped. Every number above traces to the seven sources listed below.

See also

Sources