Quantum Chip Cooling Methods: Cryogenic and Advanced Techniques

Quantum chips – especially those based on superconducting circuits – must be operated at ultra-low temperatures to function correctly. Cooling is crucial for reducing thermal noise and enabling quantum coherence. As one recent study notes, “state-of-the-art superconducting microwave qubits are cooled to extremely low temperatures to avoid sources of decoherence”. In practice this means millikelvin temperatures (<<0.1 K) are needed so that thermal excitations are negligible compared to qubit energy gaps.

For example, a 5 GHz superconducting qubit has an energy of ~20 µeV, so k_BT ≪ 20 µeV (T ≪ 250 mK) is required to keep it in its ground state almost all the time. Operating at such cryogenic temperatures suppresses thermal noise, residual quasiparticles, and blackbody photons that would otherwise disrupt quantum superposition. Indeed, experiments show that many device loss mechanisms saturate around 50 mK under normal conditions; immersing circuits in superfluid helium has been used to push effective bath temperatures even lower and break through that floor.

Quantum decoherence is especially sensitive to thermal effects. Residual heat or stray photons can create quasiparticles, activate two-level defects, and cause phase errors. For instance, Danilov et al. report that immersing a superconducting resonator in liquid ^3He dramatically increases its energy-relaxation rate and suppresses decoherence without adding noise. Conversely, Barends et al. demonstrated that inadequate shielding of infrared (thermal) radiation causes qubit quality factors to plummet. In their tests, only a fully “black” box-in-box enclosure eliminated any temperature dependence of the resonator quality, raising Q above 10^6. In short, even tiny thermal leaks – from stray photons or imperfect wiring – can reduce coherence. Thus, maintaining cryogenic conditions (via advanced cryogenics) is essential for preserving the fragile quantum hardware state.

Cryogenic Cooling: Dilution Refrigerators and Classic Methods

The workhorse of quantum chip cooling is the dilution refrigerator (DR). This system exploits a liquid helium mixture to reach millikelvin temperatures. A typical DR contains stages at ~4 K, 1 K, and a mixing chamber at ~10–20 mK. Below about 0.87 K, a circulating ^3He–^4He mixture phase-separates into a ^3He-rich (concentrated) phase and a ^3He-poor (dilute) phase. Continuously pumping ^3He atoms from the concentrated phase into the dilute phase absorbs heat, because the ^3He atoms require energy to mix. In effect, the system’s “heat of mixing” provides refrigeration power. In practice, dilution fridges routinely reach below 10 mK, with record base temperatures near 2 mK. Modern DR units use closed-cycle cryogenics (no liquid cryogens) by employing mechanical precooling.

Because dilution refrigerators are complex and large, much effort goes into their ancillary components. Early systems required liquid helium and nitrogen baths, but most current fridges are dry (cryogen-free). A key innovation is the use of cryocoolers (mechanical refrigerators) to replace the liquid-helium bath. For example, two-stage Gifford-McMahon (GM) or pulse-tube refrigerators provide ~4 K and ~50 K stages. Pulse-tube coolers are especially popular because they have no moving parts at the cold end, greatly reducing vibration (critical for sensitive qubits). In a modern setup, a cryocooler precools the incoming ^3He gas, and then a series of heat exchangers and pumping stages achieve the final mK temperatures. The diagram below illustrates a two-stage pulse-tube cooler, which often serves as the workhorse pre-cooler for superconducting qubit platforms.

A two-stage pulse-tube cryocooler – commonly used to precool dilution refrigerators. By oscillating helium gas and using no cold moving parts, pulse tubes reach ~3–4 K with minimal vibration (pulse tubes have been used for decades to reach space-like temperatures).

Beyond dilution fridges, other traditional cryogenic methods have historical importance:

• Adiabatic Demagnetization Refrigeration (ADR): This magnetic cooling technique uses the magnetocaloric effect. By magnetizing and demagnetizing a suitable salt or alloy, it achieves extremely low (µK) temperatures. ADR is widely used in astronomy detectors and fundamental physics, and has been explored for quantum experiments.
• He-3 Sorption Fridges: Adsorption of ^3He onto charcoal can cool to ~300 mK. Such sorption units are small and vibration-free, useful for pre-cooling or specialized applications. However, their cooling power is limited (µW range) and they require recycling of the charcoal pump.

Each method comes with tradeoffs. Dilution fridges offer continuous operation and µW cooling at 10 mK, but are bulky and energy-intensive (requiring ~kW of electrical power to run). Pulse-tube cryocoolers remove consumables but introduce mechanical vibration (though modern designs minimize it). ADR provides low µK temperatures without ^3He circulation, but needs cycling of magnetic fields. In practice, superconducting qubit labs rely mainly on dry dilution refrigerators with heavy vibration isolation and extensive electromagnetic shielding.

Emerging Cooling Techniques

To go beyond traditional cryogenics, researchers are developing novel cooling methods at and below millikelvin. Notable examples include:

Optical (Laser) Refrigeration: In this approach, specially doped materials are cooled by anti-Stokes fluorescence. Shining a tuned laser causes certain solids (e.g. rare-earth-doped crystals or semiconductors) to absorb lower-energy photons and re-emit higher-energy photons, removing heat. Chang and Zhang describe optical refrigeration as “an emerging research field for cooling of chips,” noting its advantages of small size, no mechanical vibration, [and] no refrigerant. Experiments have already achieved net cooling of semiconductor devices (e.g. cooling a chip by tens of kelvin from room temperature). Such laser-cooling could, in principle, be applied locally to remove hotspots or cool specific quantum devices on-chip, supplementing the cryostat.

On-chip Quantum Refrigerators: Inspired by quantum thermodynamics, researchers have built quantum absorption refrigerators from superconducting circuits. A recent Nature Physics work demonstrated an autonomous three-body circuit that cools a target transmon qubit below the base temperature of any single bath, effectively “resetting” the qubit. Driven purely by engineered thermal gradients and without active control, this tiny refrigerator lowered the qubit’s effective temperature to ~22 mK, significantly below the ~50 mK plateau of standard setups. This proof-of-concept shows that quantum systems can cool themselves using their inherent quantum interactions, a technique that could aid qubit initialization or error correction in future architectures.

Superfluid Helium Immersion: Another innovative method is immersing the chip in liquid ^3He (or superfluid ^4He). Experiments by Danilov et al. immersed a superconducting resonator in liquid ^3He and found that almost all thermal constraints disappeared – the device parameters kept improving continuously down to sub-millikelvin temperatures. The ^3He acted as an extremely effective heat sink for the qubit’s environment, increasing the relaxation of environmental modes by ~10^3 without adding noise. By exploiting the extraordinary thermal conductivity and low viscosity of superfluid helium, this method can suppress decoherence “bottlenecks” that limit conventional dilution refrigeration. (Google researchers have similarly developed “He-3 immersion” cooling for large-scale chips.)

These emerging techniques are still largely experimental. Optical refrigerators require exotic materials and high optical power, and on-chip quantum coolers are novel and fragile. Nonetheless, they point to a future where auxiliary cooling mechanisms complement the base cryostat. For instance, one can imagine lasers stabilizing hot spots, or local refrigerators on multi-qubit chips to manage heat from readout.

Phononic and Photonic Heat Transport Control

Beyond bulk refrigeration, engineering at the material and architectural level can improve thermal management. Two key fronts are phonon control and photon control:

Phononic Engineering: Phonons (lattice vibrations) are the primary carriers of heat in solids. In quantum chips, unwanted phonon modes can carry heat into qubit structures or couple qubits to substrate. Researchers are exploring phononic bandgaps and acoustic metamaterials to block or redirect phonon heat flow. For example, Kitzman et al. demonstrated coupling a superconducting qubit to a surface-acoustic-wave (SAW) phonon bath. By shaping the phonon density of states, they could stabilize certain qubit states via engineered dissipation. Other approaches include using nanofabricated phononic crystals or suspended structures (e.g. membranes, nanowires) that greatly reduce thermal conductivity to the chip. In extreme cases, device membranes can be thermally isolated from the substrate except via engineered anchors, making the qubit almost “float” thermally. Such phonon engineering reduces heat load and can even be used as a resource for state control.

Photonic (Radiative) Shielding: Even at cryogenic temperatures, photons (especially in the far-infrared/terahertz range) can cause heating and quasiparticle generation. Unintended blackbody radiation from warmer stages can leak into the qubit environment. To combat this, dilutions fridges use nested radiation shields, infrared filters, and absorptive coatings. In practice, experiments pack qubits in a box-in-box enclosure with blackbody coatings and low-pass filters on all lines. Barends et al. found that without a properly coated shield, qubit lifetimes (T1) fell sharply as the mixing chamber warmed. Only after adding successive metallic cans with absorbing coatings did the resonator Q become independent of fridge temperature. Modern quantum cryostats also include thermalization pads, cryogenic attenuators, and Eccosorb filters on each signal line to ensure no stray photon reaches the device. In short, photonic heat transport is controlled by multi-stage filtering and blackbody radiation traps, preserving the low-photon environment needed for coherence.

Materials and Thermal Interface Innovations

Materials choice and thermal interfaces are critical at cryo temperatures. The thermal conductivity of common materials changes drastically when cooled: for example, bulk silicon’s thermal conductivity rises by ~10× from 300 K to 4 K. This can be a double-edged sword. On one hand, high conductivity substrates (sapphire, high-purity Si) help spread heat away from hotspots. On the other, wiring and dielectrics can become bottlenecks if their conductivity drops. Material scientists are exploring specialized substrates (like single-crystal diamond or AlN) that combine good thermal conduction with microwave properties.

Thermal interface materials (TIMs) and adhesives also matter. At millikelvin, even a thin grease or epoxy layer can dominate thermal resistance. Quantum labs often use indium solder or gold films for mounting chips, and gold-plated copper as heat spreaders. New low-temperature solders and compliant metal interfaces (e.g. gold-indium alloys) are being developed to improve chip-to-cold-plate contact. Similarly, composite materials such as graphene or carbon nanotube films have been investigated as heat-spreading layers on fridge stages. While academic data is still emerging, the trend is clear: every interface (wire bond, glue, clamp) must be optimized for both thermal conductance and low magnetic noise.

In superconducting circuits, wiring materials also affect thermal budgets. The choice of coaxial cables (stainless steel, CuNi, NbTi) balances cryogenic thermal conductance with microwave loss. For example, Abobeih et al. quantify the passive heat load of cable trees made of stainless steel and NbTi coax (used because they conduct little heat). They show how every meter of cable and every attenuator adds heat to the base stage. This has led to high-conductivity thermal anchors: large copper blocks where many attenuators and cables are clamped at each stage. By “engineering dissipation” in this way, one can keep the chip thermally anchored to the mixing chamber while filtering out room-temperature noise.

Integration and Scalable Cryogenic Design

Building a lab-scale quantum computer involves not just cooling the chip, but integrating it with classical electronics. As designs scale to hundreds or thousands of qubits, the cryogenic integration challenge becomes enormous. Key considerations include:

Wiring and I/O Lines: Every qubit needs control and readout lines. Abobeih et al. demonstrate that operating ~50 qubits at 14 mK requires on the order of 25–50 coaxial cables into the fridge. Each line brings heat into the system. Passive conduction through the cable and active dissipation in attenuators can easily exceed tens of microwatts. Thus, cryogenic setups use extensive filtering/attenuation: attenuators at each thermal stage absorb heat from incoming microwave pulses, and add little backflow to the lower stages. This “thermal anchoring” is a standard practice: room-temperature signals are stepped down in power and thermalized at 50 K, 4 K, still (~0.8 K), and mixing chamber stages. Advanced devices like the QFilter-II can cool electrons in multiple signal lines, ensuring effective electron temperatures just 5–15 mK above the mixing chamber.

Modular Cryostat Architecture: Traditional single-chamber fridges may not scale. IBM’s Project Goldeneye exemplifies a modular approach. Goldeneye is the largest dilution cryostat ever built (by volume) to date. It was designed as a proof-of-concept to study scaling limits: it increases the experiment volume, number of I/O ports, and cooling power beyond commercial units. Although not (yet) used for a production quantum processor, Goldeneye showed how a small engineering team can push cryogenic capacity by building a “super-fridge” (see image below). Lessons from such projects include the importance of staging (multiple separate dilution loops), vibration isolation for larger cold plates, and modular I/O panels that can handle hundreds of lines.

The IBM Goldeneye super-fridge in construction. This massive dilution refrigerator was designed to explore the limits of scalability for quantum computers (scaling to millions of qubits). Goldeneye addresses the usual constraints of commercial fridges – experiment volume, number of I/O lines, and cooling power – by essentially building a fridge much larger than normal.

Thermal Budget and Cooling Power: Every quantum computer has a total cooling budget. The cold plate at ~10 mK might have only tens of microwatts of cooling power, so designers must account for it precisely. Tools like cryo-capacity mapping are being developed to predict how a given wiring harness and payload will warm each stage. Efficient cryostat design often uses multiple refrigeration units in parallel, or separate cryostats for different chiplets. For example, some proposals divide a large processor across several 10 mK stages connected by microwave links.

Integration thus becomes a co-design problem: the quantum chip architecture must take into account what is physically coolable. For instance, superconducting qubit arrays may be split across multiple mixing chambers to keep wiring density manageable. Chip layouts may include on-chip heat spreaders and cold grounds to distribute heat. All control electronics (DACs, ADCs) remain at room temperature today, but future systems may put some classical control at intermediate cryo stages (4–100 K) to reduce wiring – a trend already seen in some cryo-electronics research.

Challenges in Scaling and Future Outlook

Scaling cryogenic cooling to large quantum computers faces major challenges. One obvious issue is energy and infrastructure: a dilution refrigerator consumes on the order of tens of kilowatts of wall-plug power. (Estimates suggest a full-scale quantum server could require megawatts.) This raises sustainability concerns. It also means heat from higher-temperature stages must be rejected, placing demands on laboratory cooling systems.

Another challenge is helium usage. The global supply of ^3He is limited, and ^4He is a non-renewable resource. Many modern fridges recycle ^4He, but future technologies that can work with less or even no liquid helium (e.g. all-cryocooler designs, optical refrigeration) will be valuable.

Quantum hardware itself may evolve to ease cooling requirements. Some qubit platforms operate at higher temperatures. For example, spin qubits in silicon or germanium can have gaps large enough to work at 1–4 K, greatly relaxing cryogenics. Diamond nitrogen-vacancy qubits can function at cryogenic but near-room temperatures. Photonic qubits, used by companies like PsiQuantum, require only modest cooling (e.g. 4 K) or even chip-scale Peltier modules instead of mK fridge. As these technologies mature, they could reduce the reliance on heavy-duty dilution refrigeration.

Finally, innovation will continue on the refrigeration side. Novel cryocooler cycles, pulse-tube improvements, and even hybrid techniques (e.g. combining laser cooling with mechanical coolers) are under development. Materials science will bring better thermal interface materials and metamaterials for heat isolation. Integrated nanofabricated coolers or microscale refrigeration (using electronic NIS junctions, semiconducting 2DEGs, etc. to pump heat) are active research areas. Even in the conventional realm, companies are working on more compact, efficient DR designs and better cryogenic electronics.

In summary, cryogenics is the cold engine of quantum computing. Maintaining ultra-low temperatures is fundamental for quantum coherence and low thermal noise. As quantum processors grow, engineering ever-larger and more efficient cooling systems – from dilution fridges to photonic and phononic thermal management – will be a cornerstone of quantum computer design. The interplay between quantum chip architecture and cryogenic technology will shape the roadmap for achieving scalable, fault-tolerant quantum hardware.