Carnot batteries: Science and technology
How a heat pump and an engine make a battery
What this tab covers The physics and chemistry: how the energy is stored, which materials it needs (scarcity, toxicity, recyclability) and what makes it age.A Carnot battery uses electricity to drive a heat pump that makes a hot store and a cold store, then runs a heat engine between them to make electricity again. It uses common materials, but the efficiency of real machines is still being proven.
Mechanism
FavourablePumped thermal electricity storage (PTES) charges by running a heat pump: it moves heat from a cold store into a hot store, widening the temperature gap. To discharge, a heat engine runs across that gap. The reference design from the National Laboratory of the Rockies (NLR, formerly NREL) keeps nitrate salt between 300 and about 557 C and methanol between about minus 41 and 29 C, with nitrogen as the working gas. Brayton (gas) designs and Rankine (steam or CO2) designs differ in fluid and temperature; Malta uses molten salt and hexane.124
NLR reference hot store nitrate salt, 300 to 556.8 C4 NLR reference cold store methanol, minus 41.4 to 29.1 C4 Energy density, Brayton / Rankine PTES 10-30 / 3-15 kWh per m31Material scarcity and toxicity
FavourableThe bill of materials is steel, insulation, nitrate salt or gravel, silica sand or rock, and standard turbomachinery, with no critical minerals. Some designs use flammable cold-store fluids, methanol in NLR's design or hexane in Malta's, and molten nitrate salt is corrosive. Silica particle stores avoid both liquids.234
Degradation and ageing
MixedAgeing resembles a power plant more than a battery: turbine and compressor wear, heat-exchanger fouling, corrosion and thermal cycling of tanks, rather than capacity fade. Developers expect long lives, for example 20 years for one supercritical CO2 design. No multi-year field data exist: the largest electricity-to-electricity unit found, a 1 MW, 4 MWh pilot in Beijing, was commissioned in December 2025 after only seven full cycles.29
Expected life (Peregrine design, company claim) 20 years2Variants compared139
| Variant | Round-trip efficiency range | Capital cost per kW | Capital cost per kWh |
|---|---|---|---|
| Brayton PTES (gas cycle) | 52-70% (mean 61.6%) | USD 2,000-4,000 | USD 50-1,500 |
| Rankine PTES (steam or CO2) | 45-65% | USD 500-8,000 | USD 250-1,000 |
| Air Brayton with molten salt (SPIC pilot) | over 65% claimed (company) | not given | not given |
| Silica-particle PTES (NLR model) | 66% modelled | not given | USD 245 at 10 h; 38 at 100 h |
| Liquid air, for comparison | 40-60% | USD 700-3,000 | USD 400-800 |
Sources
- Carnot Battery development: A review on system performance, applications and commercial state-of-the-art (Vecchi et al., Journal of Energy Storage 55, 105782) University of Birmingham repository, 2022
- Review of Carnot Battery Technology Commercial Development (Novotny et al., Energies 15(2), 647) MDPI, 2022
- Advancing Pumped Thermal Energy Storage Performance and Cost Using Silica Storage Media (Applied Energy 387, 125567) National Laboratory of the Rockies (formerly NREL), 2025
- Design-Point Techno-Economics of Brayton Cycle PTES for Combined Heat and Power (Neises and McTigue, GTP-24-1410) NREL (now National Laboratory of the Rockies) / OSTI, 2024
- World-first 560 C Carnot battery unveiled as SPIC launches ultra-high-temperature heat storage (news report of company claims) ESS News, 2026