Efficient roller-driven elastocaloric refrigerator

The study addresses the need for higher-efficiency, low-emission cooling technologies in light of global commitments to reduce cooling-sector emissions. Conventional gas-phase refrigerants face environmental, safety, and regulatory challenges. Elastocaloric (eC) cooling, which leverages stress-induced solid-state phase transformations in shape memory alloys (SMAs), offers large temperature changes and practical system integration. However, many eC prototypes lack efficient work recovery: antagonistic SMA pairs recover only potential energy and suffer irrecoverable unloading work, while continuously rotary systems face friction losses whose net benefit is uncertain. Concurrently, materials advances (e.g., TiNiCu with reduced hysteresis and improved fatigue) have largely been limited to small-scale formats with scale-up challenges. The research question is whether combining materials-level improvements (nanocrystalline TiNiCu with favorable B2–B19 transformation) and a novel low-friction, roller-driven mechanism that can recover kinetic energy can significantly improve system efficiency (COP) and practical performance of macro-scale, fluid-free eC refrigerators.

Over twenty eC prototypes have been reported, with only a subset implementing work recovery. Existing schemes fall into two categories: (1) antagonistic SMA pairs that primarily recover potential energy but exhibit irrecoverable unloading work due to force mismatch at the start of unloading; (2) rotary designs that can in principle recover potential and kinetic energy continuously but may suffer from nontrivial friction, casting doubt on net gains. Materials development has focused on reducing hysteresis, enhancing reversibility, and lowering driving stress. TiNiCu alloys have shown reduced hysteresis and improved fatigue, often demonstrated at lab scale (e.g., films), but translation to macro-scale devices has been limited by scalability and processability. These gaps motivate a system architecture that allows kinetic energy recovery within a cyclic operation while leveraging scalable, high-efficiency TiNiCu ribbons.

System design: The roller-driven mechanism mounts two SMA ribbons vertically on a die-steel roller. The roller, driven via a motor and gearbox, periodically rotates between two end positions. Each ribbon is clamped to the roller at one end and to a linear sliding carriage at the other, ensuring uniaxial loading. The two ribbons operate in antiphase. Heat transfer is achieved by direct contact: at one end position, one SMA is fully loaded and contacts the heat sink (rejecting heat), while the other is fully unloaded and contacts the heat source (absorbing heat). An intermediate heat exchanger (IHX) relays heat between stages, enabling a temperature span larger than the material’s single-stage adiabatic temperature change. Kinetic energy recovery: During reversal and acceleration of the roller, unloading torque from one SMA accelerates the rotating assembly (roller, gearbox, rotor), storing kinetic energy and reducing net input work. This design seeks to minimize friction while retaining kinetic energy recovery within a cyclic operation. Materials synthesis and processing: TiNiCu ribbons were produced from a 7 kg ingot with nominal composition Ti49.2Ni44.8Cu6 (at.%). Processing steps included induction melting, forging, hot rolling to 0.85 mm, annealing at 900 °C for 0.6 ks, cold rolling to 0.64 mm (25% reduction), and low-temperature annealing at 350 °C for 1.2 ks to achieve a nanocrystalline structure. Commercial NiTi (Ni50.8Ti49.2) sheets were cut into ribbons for baseline comparison. Materials characterization: TiNiCu microstructure and composition were assessed by TEM/EDS, revealing nanocrystalline grains (~50 nm) and composition near nominal. In-situ XRD identified a B2→B19 martensitic transformation (MT), contrasting with NiTi’s B2→B19′. DSC was performed (±10 °C min−1). Mechanical tests used dog-bone specimens on an Instron 5969; training involved multiple loading–unloading cycles (~10−3 s−1) to stabilize superelasticity. Quasi-isothermal and quasi-adiabatic responses were probed at strain rates 0.0005 and 0.04 s−1; temperature changes were measured via welded thermocouple and IR camera. Prototype testing: The two-stage device used ~220 mm ribbons (120 mm heat exchange length), ~0.6 mm thick. Temperatures of heat sink/source/IHX were measured by T-type thermocouples; torque and power by dedicated meters. Operating schedules typically used 0.17 Hz (0.9 s loading, 2 s heat rejection, 0.9 s unloading, 2 s cooling). Load-free temperature spans were recorded to steady state (~50 cycles). Cooling power at zero span was determined by energy balance with a calibrated film heater on the source block. Input electric power was time-averaged over loading periods. A dynamic system model, validated by experiments, explored the impact of SMA mass ratio between high- and low-temperature stages on performance. Work recovery and COP calculations: Work recovery efficiency was defined as recovered unloading work divided by total unloading work, with a theoretical limit (without kinetic energy recovery) computed from stress–strain loops at matched conditions. Measured work recovery accounted for kinetic energy contributions via power/torque data. Material-level COP (COPmat) at zero temperature span was computed from entropy change ΔS, ΔTad, and entropy generation Sgen from isothermal loops. System-level COP (COPsys) equaled measured cooling power divided by input electric power; a baseline COP without work recovery was estimated from separate single-ribbon loading powers.

• Materials advances: TiNiCu exhibited nanocrystalline grains (~50 nm) and a B2–B19 MT with improved lattice compatibility, reducing thermal and stress hysteresis relative to commercial NiTi. After training, TiNiCu showed uniform temperature evolution without Lüders bands. Under quasi-adiabatic loading at 0.04 s−1, TiNiCu achieved ΔTad ≈ 14.2 °C (loading) and 11.3 °C (unloading). Compared to NiTi, at ~4% strain in-device, TiNiCu delivered 25% higher ΔTad and a 125% higher COPmat.
• Temperature span: Two-stage cascading increased load-free span by 82% (NiTi) and 85% (TiNiCu) versus single-stage at 0.17 Hz. Achieved spans: 19.8 K (NiTi) and 25.4 K (TiNiCu). The 25% ΔTad increase translated to ~28% larger system span (and ~26% in single-stage tests), indicating effective system integration of materials gains.
• Operating frequency: Optimal around 0.17 Hz; ribbons provided improved heat transfer versus wire-based systems, with reduced sensitivity of span to frequency (≤1.5 K over 0.18 Hz range).
• Pull-down and refrigeration: From 21 °C ambient, with TiNiCu the source cooled to 10 °C in 154 s (49% faster than NiTi) and reached 7.7 °C after 561 s (17.8 K below sink; meets wine storage requirement). A single ribbon could reach the target at 21 °C, but at 25 °C ambient the two-stage was necessary due to limited single-stage span (~10.9 °C).
• Cooling power at zero span: Optimal frequency also 0.17 Hz. Two TiNiCu ribbons achieved 4.5 W vs 3.0 W for NiTi (~50% improvement). Refrigeration span under load was 7.6 °C lower than load-free due to ambient rejection. Cascade temperature difference increased with load, limiting maximum power compared to doubling single-stage capacity.
• Mass ratio effects: Simulations (validated dynamically) showed weak influence on span (22.1 K at mass ratio 30, +11.6% vs 1:1) but significant impact on power: optimal mass ratio (SMA#2:SMA#1) ≈ 9 yielded 6.7 W vs 3.1 W at 1:1, by alleviating IHX-to-SMA#2 heat flux mismatch due to hysteresis.
• Work recovery: Time-averaged actuator input power with two TiNiCu ribbons was reduced by 44% relative to loading each individually (36% reduction for NiTi). Theoretical work recovery efficiencies (no kinetic energy recovery) were ~65.6–67.2% for both materials; measured efficiencies were higher: 78.1% (TiNiCu) and 72.1% (NiTi), evidencing kinetic energy recovery via angular momentum. Estimated kinetic energy per cycle to accelerate rotating parts was ~0.72 J, matching the ~0.8 J difference between measured and theoretical unloading work.
• Efficiency metrics: System COP with work recovery at zero span was 1.27 (TiNiCu) vs 0.70 baseline without work recovery; TiNiCu’s COPsys was 35% higher than NiTi’s. Materials COPs: 11.5 (TiNiCu) and 5.1 (NiTi). Modeling suggests improving SMA–heat exchanger heat transfer could raise COPsys toward ~2.5.

The roller-driven mechanism enables partial recovery of kinetic energy during SMA unloading, overcoming the irrecoverable work inherent to antagonistic reciprocating designs and mitigating the friction penalties of continuous rotation. Coupled with a nanocrystalline TiNiCu alloy exhibiting reduced hysteresis and improved ΔTad via a B2–B19 transformation, the system converts materials-level gains into system-level performance improvements: larger temperature span, higher cooling power, and significantly higher COPsys. The measured work recovery exceeding the theoretical limit (without kinetic contribution) validates the kinetic energy recovery concept and quantifies the energy stored in rotating components. The two-stage cascade allows spans exceeding single-ribbon ΔTad, and analysis shows that asymmetric mass allocation can boost cooling power by addressing hysteresis-driven heat flux imbalances, though it risks torque asymmetry that could degrade work recovery. A 180° phase difference between SMAs had negligible impact on span and power in simulations, supporting flexibility in timing. Overall, the approach advances practical eC refrigeration efficiency while revealing tradeoffs among angular momentum (weight/cost), cascade temperature differences, and mechanical recovery pathways.

This work demonstrates an efficient, fluid-free elastocaloric refrigerator that synergizes a nanocrystalline TiNiCu ribbon (B2–B19 transformation, reduced hysteresis) with a roller-driven, antiphase mechanism that recovers kinetic energy. The system achieved a 25.4 K load-free temperature span, rapid pull-down to 7.7 °C, 4.5 W cooling power at zero span, work recovery efficiency up to 78% (surpassing theoretical limits without kinetic energy recovery), and a system COP of 1.27—about double a baseline without work recovery. Materials COP reached 11.5 for TiNiCu, indicating headroom for system-level improvements. Future work should: (1) enhance SMA–heat exchanger heat transfer and reduce thermal contact resistance to approach COPsys ≈ 2.5; (2) design motion paths enabling continuous rotation with adequate quasi-stationary heat exchange duration; (3) optimize asymmetric mass ratios while preserving high unloading efficiency; and (4) further scale and standardize TiNiCu processing for macro-scale devices.

• System COP remains well below materials COP, indicating losses from imperfect heat transfer, thermal contact resistance, and mechanical inefficiencies.
• Cascade temperature differences inherently neutralize part of the elastocaloric effect, especially at higher loads, limiting maximum cooling power relative to the sum of single-stage capacities.
• Increasing angular momentum can raise work recovery but adds weight, cost, and potential inertia-related drawbacks.
• Asymmetric mass ratios that improve power may introduce torque imbalance and degrade work recovery efficiency.
• TiNiCu processing and scalability, while demonstrated at kilogram scale, still require further maturation for widespread adoption.
• Reported performance is under specific frequencies and boundary conditions; generalizability to varied applications may require additional optimization.