Joule Heist search

topics · updated 2026-09-25

Geothermal-Driven Sorption Cooling — Brine as Drive Heat for Absorption and Adsorption Chillers

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

Geothermal brine has driven absorption chillers in the field since 1980, but the record is thin and mixed: Oregon Tech's 1,095 kW LiBr unit on 89 °C water delivered half its nameplate at triple the heat-side flow and lost to an electric chiller in 1998; Chena Hot Springs makes −29 °C brine from 74 °C water only because it rejects to a 4 °C river; Szeged's 2025 pilot is justified by reinjection pumping savings and is still at design stage. Design rules from the geothermal-specific literature: single-effect LiBr/H2O needs ≥90 °C (COP 0.73–0.87 at 90–120 °C, flow-insensitive once the exchanger is sized), NH3/H2O ≥90 °C at COP ~0.6 (half-effect ≥80 °C at 0.3), double-effect ~1.3 only above ~140 °C. Cascading power → cooling → heat on one stream (109→99→89→79 °C) is the economic architecture (46% utilization, 1.7-year payback in Rubio-Maya's Mexico study). Against an electric chiller fed by the same plant's electricity, heat-driven cooling wins only when the heat is ≥90 °C, would otherwise be reinjected, and runs at high flat utilization; NREL's Tuttle study on 87 °C brine found it consumed ~5.6× the input energy and cost $74/MWh versus $42/MWh for heat. No geothermal-driven adsorption chiller has been studied; no data center is cooled this way.

Geothermal brine is the most obviously “free” drive heat on Earth: it arrives hot, around the clock, with no combustion and no collector field. Yet in forty-five years only a handful of chillers have ever run on it, and the best-known one was scrapped. This article assembles the design rules, the real installations, the cascade architecture that makes the economics work, and the honest comparison against the default — turning the same heat into electricity and running an electric chiller. It is the bridge between the wiki’s sorption articles (Absorption, Adsorption) and the two geothermal articles (Ground-Source, EGS).

Why the field record is so thin

The Geo-Heat Center put it plainly in 2006: no absorption chillers ran on US geothermal sites between 1989 and 2004, because the ~110 °C (230 °F) classically needed for a cost-effective air-cooled machine “encroached on temperature ranges which could similarly be used for power generation,” and because unique systems bring maintenance problems. Any brine hot enough to drive a chiller well is hot enough to sell as electricity, and Francisco-Hernandez et al.’s 2025 review of installed cases finds the mean drive temperature of geothermal building-cooling systems is ~110 °C — right on that boundary. The technology is not exotic (38% LiBr/H₂O, 46% NH₃/H₂O, mean capacity ~1,280 kW at ~35 kg/s of geofluid, mean payback 5.6 years across published cases; food-refrigeration cases are far smaller, ~10 kW at 90–100 °C and COP ~0.6); the niche is narrow. Cooling remains one of the two smallest US geothermal direct-use categories.

Design rules: what the brine temperature buys

Compiled from the geothermal-specific modelling literature and the manufacturer data used in cascade studies:

CycleMinimum driveCOPNotes
LiBr/H₂O single-effect≥ 90 °C0.73 at 100 °C; up to ~0.87 at 120 °C (SHX ε 0.9); 0.76–0.80 across 80–105 °C brineCOP falls above the optimum as heat input grows faster than cooling. Crystallization risk rises at high absorber UA (limits ~1.9 kW/K at 90 °C, 2.5 at 110 °C). Cooling above 0 °C only. Exergy COP 0.50–0.69, better at higher chilled-water temperature.
LiCl/H₂O single-effect≥ 90 °C0.75–0.79Within 1–2% of LiBr; viable alternative pair.
NH₃/H₂O single-effect≥ 90 °C~0.6For sub-zero refrigeration and ice. 10 °C generator ΔT typical.
NH₃/H₂O half-effect / double-lift≥ 80 °C (74 °C proven with a 4 °C sink)~0.3Trades COP for lift; Chena’s three-pressure double-lift unit is this class.
LiBr/H₂O double-effect> ~140 °C~1.3Only one geothermal parametric study found (475 kW). Competes directly with ORC for the same brine.
Two single-effect LiBr in series90–100 °C then 80–85 °C0.74 and 0.60; cascade 0.67UWA deep-well campus study (2.5–3 km wells, 1–3 MW); payback 11–13 yr.

Three second-order rules matter as much as the table:

  1. Heat-side flow, not temperature, is what a low-grade resource costs you. Once the heat exchanger is sized, brine mass flow (1 vs 50 kg/s) has almost no effect on COP or capacity; drive temperature and HX area are the levers. But at low ΔT the flow per ton explodes — Oregon Tech pumped 685 gpm for 150 tons (~4.6 gpm/ton at 11 °C ΔT), roughly three times the heat-side flow of a chiller at design temperature, and returned most of the pumped heat to disposal unused. That flow is pumping power and reinjection cost.
  2. A cold sink is worth more than a hot source. Chena’s double-lift ammonia unit delivers −29 °C brine from 74 °C spring water because it rejects to 4 °C river water; when the creek warmed to ~7 °C on summer afternoons, coincident with peak load, output fell, and the owners planned a 3.9 °C shallow cold-water well instead. Oregon Tech’s 89 °C source underperformed, and Tuttle’s ~60 °C scenarios were excluded from cooling entirely. The working pair matters here: the YORK whitepaper warns that LiBr machines need entering condenser water of at least ~20 °C or they crystallize, so the “colder is better” rule applies without limit only to ammonia-water; for LiBr the sink should be cool but not cold. Either way, it is the same argument the wiki makes for earth-coupled rejection in Geothermal / Ground-Source — and it is why a single-effect machine rated at 116 °C delivered ~48% of nameplate on 89 °C water.
  3. Materials. Direct brine in the generator ran for 18 years at OIT in 90-10 cupro-nickel without tube failure, but corrosion was still cited at retirement, and H₂S and ammonia in the same campus water destroyed solder, thinned copper and failed brass and stainless valve trim; the campus moved to isolating plate heat exchangers. Chena used stainless on every surface touching spring water, river water or brine. Isolate the loop.

The installations

Oregon Institute of Technology, Klamath Falls (1980–1998). A nominal 312-ton (1,095 kW) LiBr/water chiller, installed for $171,300, fed 89 °C well water directly into its generator tubes to base-load five buildings (25,800 m²) at 7 °C chilled water, with an existing electric centrifugal for peaking. Rated for 116 °C, it produced 150 tons (526 kW); it needed 37.8 L/s of brine while extracting only 11 °C; the generator copper corroded. It was replaced by an electric centrifugal in 1998 “due to the low efficiency and high water usage.” The canonical cautionary tale — with the caveat that a 1980 machine is not a modern low-temperature LiBr unit (Yazaki-class chillers are rated at 70–95 °C hot water; see Absorption Cooling).

Chena Hot Springs, Alaska (2005–present). A 4.26 kW custom three-pressure ammonia-water chiller (Energy Concepts) holds an ice museum of ~1,000 t of ice at −4 °C through 32 °C summers, driven by 74 °C spring water (5.4 L/s) and rejecting to 4 °C river water (5.0 L/s), with a 7.6 m³ brine tank giving about an hour of ride-through. The spring water leaves at 63 °C and cascades to a greenhouse and pool. Whole-system electrical draw ~33 kW versus 110 kW for the diesel-powered vapor-compression backup, at ~30 ¢/kWh site power — about $400/day saved. Availability ~95%; the problems were air-handler defrost (manual, several times a day) and brine-temperature trips below −34 °C, not the cycle. The manufacturer had a precedent: a similar unit ran ~10 years at Kotzebue making fishing-fleet ice from 74 °C diesel jacket water. Chena is also the world’s lowest-temperature geothermal power plant (730 kW ORC at 73 °C) — the same cold sink enables both.

Szeged, Hungary (pilot, 2025–). On a 27-well, 76 MWth geothermal district-heating system with 95 °C production water from 2,000 m, summer heat demand collapses and return water exceeds 50 °C, raising reinjection pumping power. A 45 kW-load absorption chiller (thermal COP ~0.8) at the utility’s own headquarters, with PV for auxiliaries, is justified as a reinjection-temperature-lowering device (−5 °C) as much as a cooler. Geothermal heat at 16.7 EUR/GJ gives cooling at 0.075 EUR/kWh versus ~0.08 for conventional AC (SCOP 3 at 0.24 EUR/kWh) — parity, with pumping savings as the tie-breaker. Eight other Szeged circuits inject at ~50 °C, so it is replicable; the retrofit also requires re-insulating uninsulated heating pipes against condensation. Design stage; no operating data yet, and the paper is internally ambiguous about whether the chiller sees 95 °C production water or 40–60 °C return water — which, given the ≥90 °C rule above, is not a small ambiguity.

Others in the literature. Cerro Prieto (18 kW NH₃/H₂O), Los Azufres (10.5 kW), Las Tres Vírgenes (half-effect NH₃-LiNO₃) and Puertecitos (35 kW) in Mexico; a Jamaican geothermal milk store; Yilmaz’s Turkish NH₃-water model at 100 °C / 100 kg/s (COP 0.44, 8.8-yr payback). Ketfi et al. (2023) invert the arrangement and use geothermal wells as the sink for the absorber and condenser, finding a double-effect machine needs ~35% more ground loop.

The cascade: where geothermal cooling earns its place

The architecture that makes geothermal sorption cooling pencil is not “brine → chiller” but power → cooling → heat → reinjection on one stream, each stage taking a ~10 °C slice:

109 °C brine ─► ORC (125 kWe, η 10.6%) ─► 99 °C ─► NH3/H2O absorption (674 kW cold, 31 t/h ice) ─► 89 °C ─► direct-use HX ─► 79 °C ─► reinject

Rubio-Maya et al.’s optimized cascade for Mexico’s 276 low/mid-enthalpy sites (average 110 °C) reaches 46% overall energy utilization, NPV $10.7M and a 1.7-year simple payback (a half-effect variant on 80 °C: 337 kW, 36%, NPV $5.2M) — and its key insight is that “low ORC thermal efficiency is not a downside,” because the small-ΔT ORC leaves hotter brine for the sorption stage; the authors add that medium-enthalpy direct uses earn about three times the revenue of power alone. The same logic appears wherever it has been modelled: adding a LiBr chiller downstream of a geothermal ORC raised total energy efficiency from 9.3% to 47.3% (Ehyaei et al. 2020); El Haj Assad’s single-flash plant (250 °C production, ~13% plant efficiency) bottomed with a LiBr chiller on its 100 °C reject gives COP 0.73; Leveni and Cozzolino’s cascade ORC/absorption reaches 32.8% exergy efficiency against 35.5% for a Goswami cycle; Schifflechner et al. compared eight serial and parallel ORC-plus-chiller layouts with part-load behaviour. A caution from the review: some “geothermal CCHP” papers (e.g. Shamoushaki et al. 2021, 58–61% exergy efficiency) make their cooling with a vapor-compression loop on the ORC fluid, not sorption, so headline trigeneration efficiencies need checking. Chena is the cascade in miniature (74 → 63 °C → greenhouse). NREL’s Deep Direct-Use assessments found absorption-chiller cooling on “a high-quality geothermal resource” attractive against decentralized AC, inside an LCOH range of $13–350/MWh.

The cascade has a hard prerequisite: the brine leaving the turbine must still be ≥ 90 °C. That holds for flash plants and small low-ΔT ORCs on 110 °C hydrothermal water. It does not obviously hold for an EGS binary plant.

The EGS question: turbine or chiller?

EGS brine arrives at 175–199 °C and the ORC is designed to squeeze it — Cape Station targets 199 °C precisely for “optimal power conversion efficiency,” yielding 94 kWe per kg/s. Fervo does not publish reinjection temperature, but a utility-scale ORC optimized for electricity does not leave 90 °C on the table. So a sorption stage on an EGS plant is not free bottoming; it takes brine upstream of or parallel to the turbine and forgoes electricity. The comparison, as a wiki-derived illustration per unit of brine heat, is then:

RouteCooling delivered
ORC at 10–15% → electric chiller at COP 3.5–5~0.4–0.6 units
Single-effect absorption directly~0.7 units
Adsorption directly (60–95 °C drive)~0.5–0.6 units

Heat-driven wins on thermal accounting by a modest margin — and loses on nearly everything else: the ORC’s electricity is fungible and sells at $50–100/MWh into a PPA; the chiller’s cold is not; absorption rejects ~2.4 kW per kW of cooling versus ~1.15 for electric, roughly doubling cooling-tower duty and water in the arid basins where EGS sites are; and YORK’s own worked 500-ton example (COP 0.7 vs 6.5) shows single-effect absorption beating an electric centrifugal on energy cost only when electricity exceeds ~$0.127/kWh with steam at $4/1,000 lb — a break-even derived from the vendor’s figures, not stated by it. NREL’s Tuttle study is the quantitative floor: on 87 °C brine from repurposed 3.3 km oil wells, absorption cooling (COP 0.7–0.8) consumed 8.64 GWh/yr of heat for loads an air-cooled electric baseline met with 1.55 GWh/yr — ~5.6× the input energy — and cost $74/MWh of cold versus $42/MWh of heat; new wells pushed LCOH toward $500/MWh, and some well configurations showed rapid thermal drawdown after two years.

So the conditions under which geothermal heat belongs in a chiller are specific:

  • the brine is ≥ 90 °C after whatever power stage it feeds (flash plants, low-ΔT ORCs, or no power stage at all);
  • that heat would otherwise be reinjected unused — the summer district-heating problem (Szeged) is the clearest case;
  • the cooling load runs at high, flat utilization (Tuttle’s LCOH was $198–243/MWh with no base load versus $68–94/MWh with 0.6 MW);
  • electricity is expensive or unavailable (Chena’s 30 ¢/kWh diesel) and rejection water is cheap or cold;
  • or the system’s real product is a lower reinjection temperature, with cold as a by-product.

For data centers — the market actually buying EGS — the trade association Geothermal Rising proposed to DOE in May 2025 that geothermal heat feed absorption chillers directly, “eliminating energy waste from converting heat to electricity,” inside a power → cooling → industry → district-heat cascade; but no such installation exists, every hyperscaler geothermal deal (Google–Fervo, Meta–Sage, Meta–XGS) is an electricity PPA, Rhodium calls absorption chillers for data-center cooling “still nascent,” and YORK/Johnson Controls itself calls electric centrifugals “a sound choice for mission-critical applications such as data centers.” Treat geothermal-heat data-center cooling as proposal-stage.

Adsorption: the unstudied option

Everything above is absorption. An open-access literature search (2018–2026) found no peer-reviewed study of a geothermal-driven adsorption chiller. That is a real gap, not a verdict: silica-gel/water and zeolite/water machines run on 60–95 °C hot water (Adsorption Cooling, Commercial Adsorption Chillers), which is exactly the band where LiBr machines derate and where geothermal return water and ORC reject sit. Adsorption avoids LiBr crystallization and corrosion, tolerates variable drive temperature, and — like Chena — benefits from a cold rejection sink without the LiBr crystallization floor. What is missing is any geothermal-specific cost or field data; the thermal-COP penalty (roughly 0.5–0.6 versus 0.7) and the low specific cooling power of adsorbers count against it on a per-kg/s-of-brine basis. A geothermal district-heating utility with unused summer return water — Szeged’s 40–60 °C returns are the concrete example — is the obvious pilot host.

See also

  • EGS — where the 175–200 °C brine comes from and why it is optimized for the turbine
  • Geothermal — the shallow end: earth as the cold sink that makes low-temperature drive heat usable
  • Absorption Cooling — the cycle and vendors; LiBr crystallization and the 70–95 °C Yazaki class
  • Adsorption Cooling — the solid-sorbent alternative that fits the 60–95 °C band but has no geothermal literature
  • Commercial Adsorption Chillers — machines that could be tried on geothermal return water
  • PV vs Thermal — the same “make electricity or use the heat?” decision with sunlight instead of brine
  • Vapor-Compression Cooling — the electric-chiller baseline that OIT, Tuttle and YORK all measured absorption against
  • Comparison — COP and drive-temperature table for all cycles
  • Overview

Sources