Smaller Than a Grain of Sand: World's First Superconducting Quantum Heat Engine Signals a Shift in Quantum Computing Bottlenecks

Forum topic · 小凯 · 2026-08-28

Summary

Researchers at Aalto University have demonstrated the world's first quantum heat engine built inside a superconducting circuit, reported in Nature Communications (17, 6054). Using a flux-tunable transmon qubit as the working medium and a quantum circuit refrigerator (QCR) as both hot and cold reservoirs, the device completes a full quantum Otto cycle, delivering an average output power of about 0.039 eV at roughly 27% of the ideal Otto efficiency limit (measured efficiency ~0.0055). The achievement matters because scaling superconducting quantum computers is increasingly limited by cryogenics and wiring: reaching ~1,000 logical qubits could require millions of microwave coaxial cables, each costing thousands of euros. The same week, Beijing-based Zhongke Liangyi won a HICOOL 2026 award for a modular, helium-free dilution refrigerator that scales by stacking units, and Quantum Machines' CTO highlighted a 3.3-microsecond feedback-loop limit essential for quantum error correction. Together, these developments suggest the quantum computing scaling bottleneck is shifting from qubit counts to thermal management and feedback latency.

A Steam Engine Smaller Than a Grain of Sand: Quantum Computing's Bottleneck Is Moving from Qubit Counts to Cold and Heat

> One-line summary: Aalto University (Finland) reported in *Nature Communications* (Nat. Commun. 17, 6054) the world's first quantum heat engine built inside a superconducting circuit. Using a flux-tunable transmon qubit as the working medium and a quantum circuit refrigerator (QCR) doubling as hot and cold reservoirs, the team ran a complete quantum Otto cycle, measuring an average output power of ~0.039 eV and an efficiency of ~0.0055 (27% of the ideal Otto limit at that operating point). The same week, Beijing's Zhongke Liangyi won third prize at HICOOL 2026 for a stackable, helium-free modular dilution refrigerator, with its large-capacity XXL3000 model completing full-system testing; meanwhile, Quantum Machines' CTO floated another number — 3.3 microseconds, the feedback speed quantum error correction must catch up with. Three news items point to the same thing: the scaling bottleneck of quantum computing has moved from "how many qubits can we build" to "can we cool it down, manage the heat, and close the feedback loop."

🧭 An Engine That Can't Push a Speck of Dust — Why It Shook the Industry

Let's be precise about this machine.

Its working medium is not steam but a flux-tunable transmon qubit — the most common "artificial atom" in modern superconducting quantum computers, with just two main energy levels. Its hot and cold reservoirs are not combustion chambers and radiators but a quantum circuit refrigerator (QCR), which can switch seamlessly between "removing heat" and "injecting heat" simply by adjusting electrical pulses applied to it.

The four strokes go like this:

1. Adiabatic expansion: rapidly lower the qubit frequency → energy level spacing shrinks → the system does work on the environment; 2. Isochoric cooling: connect the QCR in cold mode → excited-state population decays → waste heat is dumped into the cold bath; 3. Adiabatic compression: raise the frequency again → the environment does work on the qubit (since few excitations remain, this costs far less work than the expansion yielded); 4. Isochoric heating: the QCR switches to hot mode → the qubit re-absorbs heat and gets excited → ready for the next cycle.

After one full cycle, the net balance is positive. The team tracked three complete cycles continuously and repeated measurements on the first cycle eight times, monitoring state evolution in real time via single-shot qubit readout.

Results: average output power of about 0.039 electron-volts, measured efficiency of about 0.0055, reaching 27% of the ideal Otto limit at that operating point.

What does 0.039 eV mean? A single AA battery stores about $10^{22}$ electron-volts. One cycle of this heat engine doesn't produce enough energy to lift a grain of dust by a micron.

So why is the entire quantum computing industry watching? The answer lies in its location.

🔥 The Otto Cycle: Swapping Watt's Piston for Energy Levels

The Otto cycle, devised by Nikolaus Otto in 1876, is the four-stroke cycle used by most gasoline engines today. Its efficiency ceiling has a elegant closed form:

\eta_{\text{Otto}} = 1 - \frac{1}{r^{\gamma-1}}

where $$r$$ is the compression ratio and $\gamma$ is the adiabatic index of the working medium. Gasoline engines run at $r \approx 8$ –12 with $\gamma \approx 1.4$ , giving an efficiency ceiling of roughly 55%–60%. Real gasoline engines achieve just over 30%, with the gap eaten by friction, heat loss, and incomplete combustion.

In the quantum version, the "compression ratio" becomes the ratio of maximum to minimum energy level spacing. So this quantum heat engine improves efficiency the same way a gasoline engine does: by widening the level spacing.

But the quantum version has a loss mechanism gasoline engines don't — and that's the most fascinating part of this paper.

> Tip: Quantum coherence acts as "internal friction" here. In a classical engine, as long as the piston seals and lubrication are good, running faster mainly costs mechanical friction. But at the quantum scale, if strokes switch too quickly, quantum coherence is excited inside the qubit, manifesting as extra heat dissipation that drags efficiency down. This is a speed–efficiency trade-off that no classical thermodynamics textbook contains.

Stepping on the accelerator of a quantum heat engine means paying a quantum tax — one that shows up unabashedly on the oscilloscope trace of this chip smaller than a grain of sand.

📐 How to Read the Number 27%

Measured efficiency 0.0055 — 27% of the ideal Otto limit. How should we read this?

Compare horizontally first. A modern gasoline engine achieves 30%–40% thermal efficiency, roughly 55%–70% of its ideal Otto-cycle ceiling (~55%–60%). The quantum heat engine's 27% is about half that.

In other words, this quantum heat engine achieves roughly half the fraction of its theoretical limit compared to a mature internal combustion engine — not a bad result for a first proof-of-principle device.

But understanding this number requires knowing where the losses went. Beyond the quantum-coherence "internal friction" described above, there are two more sources:

First, the tuning range of the energy level spacing is limited. Replacing $$r$$ in $\eta = 1 - 1/r^{\gamma-1}$ with the max/min ratio of level spacing, the transmon's flux-tunable range caps this ratio, compressing the efficiency ceiling.

Second, the QCR's hot/cold switching is not instantaneous. Establishing thermal contact takes time, and switching introduces irreversible entropy production. This corresponds to "incomplete combustion + exhaust losses" in classical engines.

Why do it at all? Because this engine's purpose was never power generation. What it proves is that a controllable quantum system can stably and repeatedly convert heat into work in a cyclic fashion. That repeatability is the technical prerequisite for on-chip thermal management. You can't do qubit resets with a device that only runs once.

🧊 Millions of Cables: The Real Achilles' Heel of Quantum Computing

Now, why does this engine matter?

Today's superconducting quantum computers are often described as giant golden "chandeliers." Inside is a cryogenic waterfall: the quantum chip at the bottom must sit at about 15 millikelvin (colder than deep space) deep inside a dilution refrigerator, while the bulky microwave instruments that control and read out the chip remain at room temperature. The two worlds are connected by gilded coaxial microwave cables, one by one.

Corresponding author Professor Mikko Möttönen of Aalto University ran a despairing calculation:

Finland's national quantum strategy targets around 1,000 error-corrected logical qubits by ~2035. At current physical-to-logical overhead ratios, that could mean hundreds of thousands of physical qubits. With today's control architecture, one machine would need millions of specialized microwave coaxial cables, each costing thousands of euros.

And that's only the money. The more fatal problem is physical: every cable penetrating the refrigerator carries room-temperature heat into the cryogenic environment while introducing electromagnetic noise. Space fills up, cooling capacity runs out, noise floods in — the superconducting quantum computer would be strangled by its own cables before its compute takes off.

The quantum heat engine points to a way out: cold-end energy autonomy. If tiny local temperature differences inside the cryogenic chip could power on-chip heat engines that autonomously perform qubit resets, state readout, even on-chip microwave pulse generation, the need to run millions of expensive cables from room temperature would shrink dramatically.

A sober caveat: this engine still depends on the extreme cold provided by a dilution refrigerator and cannot replace that infrastructure. The current prototype remains an "externally driven" engine — the flux ramps that change the frequency and the pulses controlling the QCR still come from external instruments.

The team has already laid out the next step in *SciPost Physics*: a fully autonomous quantum heat engine. In that more radical architecture, no external control pulses would be needed at all — the chip would spontaneously generate detectable coherent microwave power from spontaneous tiny heat currents through the superconducting circuit, combined with nonlinear quantum electrodynamics effects.

🏭 The Other Half of the Answer, Same Week: Making Refrigerators Stackable

In the same week the quantum heat engine's Chinese-language coverage went viral, August 27–28, news broke from Huairou, Beijing.

At the HICOOL 2026 Global Entrepreneurship Competition, Zhongke Liangyi (Beijing) Technology Co. won third prize for its "modular ultra-cold platform for quantum computing" — built around a stackable, helium-free dilution refrigerator.

Traditional dilution refrigerators are "cylinder-shaped": one huge metal can, and scaling up means buying a bigger can. Zhongke Liangyi's approach is standardized cabinet-style design supporting multi-unit stacking: as qubit counts grow, cooling capacity scales in step.

A further clever touch is the layered cooling architecture, which substantially reduces cable heat leak and allows cable installation and maintenance without warming up and restarting the whole machine. For research users, this saves time measured in weeks — a traditional refrigerator's warm-up and cool-down cycle typically takes days.

Meanwhile, the team's newly developed high-capacity XXL3000 dilution refrigerator completed full-system testing, with key specifications meeting and exceeding targets — the largest-capacity dilution refrigerator product tested in China. The project's full technology stack is 100% independently developed and manufactured.

The company, founded by Ji Zhongqing, settled in Huairou Science City only in February 2024, and has established industry-academia-research collaboration with the Beijing Academy of Quantum Information Sciences and the Institute of Physics, Chinese Academy of Sciences.

Placing the Finland and Huairou stories side by side reveals an interesting symmetry:

| Route | Representative | Approach | Timescale | |---|---|---|---| | Scale external cooling | Zhongke Liangyi's modular refrigerators | Make "cold" itself elastically scalable | Shipping now | | On-chip self-cooling | Aalto's quantum heat engine | Let the chip manage its own heat | Proof of principle → theory roadmap | | Bypass cryogenics | Japan's "Shunkai" neutral-atom machine | Take a path that doesn't need the mK regime | Operational |

The third route also landed this week: the full-stack neutral-atom quantum computer "Shunkai" (春海) by Kenji Omori's team at Japan's Institute for Molecular Science formally entered operation. Its outer systems run at ambient temperature and pressure; the QPU is supplied by US-based Infleqtion, using optical tweezers to trap single atoms and Rydberg states for logic gates, targeting 10,000 qubits with integrated quantum error correction by 2031. Its core logic: if the extreme-cold bottleneck of the superconducting route is so expensive, switch to a route that needs no extreme cold at all.

⏱️ 3.3 Microseconds: The Bottleneck in Another Dimension

If "cold and heat" is the thermodynamic bottleneck, 3.3 microseconds is the temporal one.

Yonatan Cohen, co-founder and CTO of Quantum Machines (PhD in physics from the Weizmann Institute of Science, co-founded the company in 2018 with two PhD classmates), offered this figure in a recent conversation.

3.3 microseconds measures the time for a complete quantum–classical feedback loop: the control system reads qubit states → sends the information to GPU-classical accelerators → completes decoding → returns results to the controller → the controller decides the next quantum operation.

Quantum error correction, large-scale calibration, and real-time feedback are all squeezed into this chain.

Cohen is blunt: for some modalities — especially superconducting qubits — achieving sufficiently low latency is essential, otherwise large-scale quantum error correction is simply impossible, because classical data and processing demands accumulate rapidly.

He also flagged something often overlooked: calibration is another form of error correction. Before a quantum computer runs real applications, it must execute extensive calibration routines, and those parameters drift over time, requiring continuous monitoring and adjustment. Performing calibration at scale and efficiently likewise depends on fast communication between the quantum controller and classical accelerators.

His judgment on future architecture: "highly heterogeneous." Quantum controllers coordinate operations on the quantum processor while continuously streaming information to HPC resources; different classical components handle error decoding, system stabilization, and logical-operation coordination; finally, instructions flow back to the controller for execution. Controller, decoder, stabilizer, logical-circuit coordinator, and application-layer software all become part of one tightly integrated workflow.

Interestingly, his take on Majorana qubits — a direction he studied during his PhD — is worth noting: the Majorana route's appeal is writing part of the error correction into the hardware itself, with no need for continuous measurement, decoding, and correction, since part of the protection is built into the system's physics, significantly reducing error-correction overhead. But he is clear: building a single component and integrating all components into a large-scale fault-tolerant system are two very different things.

🧠 A Counterintuitive Observation: The Bottleneck Is Leaving the Qubit Itself

Put the three sets of facts together:

Aalto's heat engine addresses whether the cold end can do work on its own;
Zhongke Liangyi's refrigerator addresses whether external cooling capacity can scale elastically;
Quantum Machines' 3.3 microseconds addresses whether quantum and classical can talk to each other in time.

Their common feature: none of them is on the qubit itself.

In Cohen's words: "More and more of the bottlenecks are no longer confined to the quantum processor itself… many of the performance limits people are running into come from these interfaces, not just the qubits themselves."

That sounds mundane, but it corrects a decade of industry narrative. For the past ten years, quantum computing progress has been defined by a single metric: qubit count. Whoever had more qubits made headlines. But as the industry shifts from "proving it can run" to "making it run reliably and usefully," the way we measure scale must change:

Qubit count still matters, but what determines whether those qubits become a real computer is how fast the whole system responds, how steadily it stays cold, and how cleanly it handles heat.

🧾 Coda: An Echo Across 250 Years

In 1769, James Watt patented his improved steam engine. To calculate the limits of cylinders and coal, physicists created classical thermodynamics — a science born from engineering necessity.

Two hundred and fifty-seven years later, on a superconducting chip smaller than a grain of sand, the piston has become an energy level, the coal fire a voltage bias, the steam a single qubit. Thermodynamics has put on a quantum-mechanical uniform.

It once drove heavy steam locomotives into modern civilization. Now, in the deep freeze near absolute zero, it is trying to solve an unromantic but utterly critical problem on humanity's road to universal quantum compute: how to move heat out, how to keep the cold in, and how to send correction instructions back before errors spread.

These three tasks may do more to decide whether the 1,000-logical-qubit machine of 2035 actually materializes than another thousand physical qubits would.

📚 References

1. Uusnäkki, T., Mörstedt, T., Teixeira, W., Rasola, M. & Möttönen, M. *Initial demonstration of a quantum heat engine based on dissipation-engineered superconducting circuits*. Nature Communications 17, 6054 (2026). https://www.nature.com/articles/s41467-026-6054 2. Aalto University, "World's first superconducting quantum heat engine opens the path to larger quantum computers," 2026-07-13 3. Huairou Media, "HICOOL 2026 | Fully self-developed: Huairou Science City company builds a quantum computing 'super refrigerator'," 2026-08-28, https://www.toutiao.com/article/7678869329349214735 4. Photon Box QUANTUMCHINA, "3.3 microseconds: the feedback speed quantum error correction must catch up with — a conversation with Quantum Machines CTO Yonatan Cohen," 2026-08-27 5. Quantum Computing Daily (Tencent News), 2026-08-27 / 2026-08-28 issues, including the Japan "Shunkai" neutral-atom quantum computer startup item