Cryogenic PCB: How Multilayer Boards Work at 11 mK
A cryogenic PCB is not a peripheral component in a quantum computer. Research published on 24 August 2026 by Oxford Quantum Circuits described microwave Purcell filters integrated directly into a multilayer board and validated on a thirty five qubit superconducting processor. The board used low loss material, each embedded filter section coupled up to nine readout resonators, and thirty five resonators were multiplexed onto six radio frequency output lines. The device was cooled to eleven millikelvin, and the median qubit relaxation time reached eighty four microseconds.
The result marks a change in role. Printed circuits in quantum systems traditionally carried signals between the processor and the surrounding electronics, while filtering occupied area on the quantum chip itself. Moving that function into the board uses the vertical dimension of the package to relieve pressure on chip area, which becomes increasingly important as qubit counts rise.
What the Experiment Demonstrated
Three outcomes matter for board manufacturers. First, a multilayer printed circuit can implement a distributed microwave function rather than only routing signals. Second, that function performed well enough to be measured on a real superconducting processor. Third, the approach scaled to thirty five resonators through multiplexing, which is the property a practical readout system needs.
Electromagnetic simulation indicated that the board level filter substantially suppressed relaxation of the qubits toward the readout port, which is precisely the loss mechanism the Purcell filter exists to prevent. The physical demonstration followed the simulation, which is the sequence that makes a result credible for design reuse.
Why Low Loss Material Matters More in the Cold
Dielectric loss is the enemy of any high frequency circuit, and in a quantum readout chain it is critical because the signal is extraordinarily weak. The board in this work used a low loss polytetrafluoroethylene based laminate chosen for its stable electrical properties, and that choice is what allows a filter implemented in copper and dielectric to behave predictably near absolute zero.
Thermal contraction is the second consideration. Materials shrink as they cool, and a multilayer stack that contracts differently from layer to layer develops stress and can change dimensions enough to shift a resonant frequency. Controlling the stack, its symmetry and its material selection is what keeps the electrical design valid at operating temperature.
Filters Inside the Board
Embedding a filter means that specific layers carry resonant structures rather than continuous planes and traces. Their geometry determines the centre frequency and bandwidth, so line width, dielectric thickness and copper thickness all become electrical parameters with tight tolerances rather than process variables with loose ones.
This is where fabrication capability directly limits design freedom. A filter structure that requires line widths and gaps within a few micrometres of target, on a material that is difficult to etch uniformly, will not reproduce across a panel unless the process is well characterised. The board becomes a component with a specified performance rather than an interconnect with a specified geometry.
Vertical Interconnect and Multiplexing
Multiplexing nine resonators onto one output requires transitions between layers that behave consistently. Each transition introduces a small discontinuity, and where several channels share a common path, differences between them show up as variation in the readout rather than as a simple loss.
Using the vertical dimension of a three layer structure allowed the filter units to be placed in an inner layer with interconnections brought to the surface. That arrangement keeps the filter close to the signal path while leaving the outer layers available for routing and shielding, and it demonstrates why designers of cryogenic hardware now treat the board as part of the microwave design.
Thermal Behaviour at Millikelvin Temperatures
At eleven millikelvin there is almost no thermal energy available, so heat conduction and heat capacity both behave differently from room temperature. Copper remains an excellent conductor, but the dielectric contributes phonon paths that can carry noise into sensitive structures, and the board itself becomes a thermal load on the refrigerator.
That makes mass and thermal anchoring engineering parameters. A board that is too heavy or too poorly anchored takes longer to cool and can introduce temperature gradients between the processor and the filtering stage. Designers therefore consider how the board mounts, where it is thermally tied and how much material is strictly necessary.
Material Behaviour When Cold
Properties that are convenient at room temperature can be unsuitable in a cryostat. Some adhesives become brittle, plated finishes can behave differently, and solder joints experience repeated thermal cycling from ambient to millikelvin temperatures. Mechanical integrity under that cycle is a reliability requirement rather than a formality.
Material selection therefore considers the full temperature range. Laminates with stable dielectric behaviour and low moisture absorption are preferred, and assembly processes may need adjustment because the thermal expansion mismatch between a low loss laminate and copper is different from that of standard FR-4.
Fabrication Challenges
Low loss materials are more difficult to process than standard laminates. They tend to be softer, they are more sensitive to moisture before lamination, and their dimensional behaviour during pressing differs, which complicates registration on a multilayer stack with tight tolerances. Surface preparation before plating also needs care to achieve adhesion without roughening the copper enough to increase loss.
Fine features add to the difficulty. Filter structures with narrow gaps demand good etch control, and where the design uses multiple laminations the accumulated movement has to stay within the tolerance of the resonant geometry. Manufacturers serving this market typically run these materials continuously rather than occasionally, because process parameters for them are not transferable from FR-4 experience.
Assembly and Handling
Cryogenic assemblies use connectors and components that must remain reliable across a very wide temperature range, and the solder joints that attach them experience the same cycling as the board. Mechanical support matters because the contraction of the assembly can load connectors in ways that do not occur at room temperature.
Handling and cleanliness also carry more weight than in ordinary electronics. Surface contamination that would be harmless at room temperature can influence loss and noise in a cryogenic environment, so assembly practices closer to those used for high reliability electronics are appropriate.
Testing Without Warming the Device
Verifying a cryogenic board is inherently difficult because the properties that matter appear only at operating temperature. Room temperature measurements of impedance and insertion loss provide a first check but do not capture the behaviour of the material when cold, and the meaningful verification happens in the dilution refrigerator.
Practical programmes therefore combine simulation, room temperature radio frequency measurements on test structures, and a limited number of full cold tests. Designing test coupons that can be measured in both conditions helps establish the relationship between warm and cold behaviour, which reduces the number of expensive cryogenic runs needed.
What This Means Beyond Quantum Computing
Embedding filters and multiplexers in the board is not a technique that stays confined to one application. The same approach appears in radio frequency front ends, phased array antennas and instrumentation, where the board does part of the signal processing instead of merely connecting components.
The common requirement is fabrication precision and material stability, which is also what advanced communication and radar boards need. Manufacturers that develop the capability for one of these applications generally find it transferable, and that is why high frequency fabrication capability is worth evaluating on the hardest product a supplier has built rather than on the easiest.
Why Quantum Hardware Needed the Board
Readout chains in superconducting systems are limited by how many signal lines can reach the processor inside a dilution refrigerator. Each line carries heat and noise into the cold stage, so reducing their number is a system level requirement rather than a packaging preference. Multiplexing on the board directly reduces that burden.
The same logic applies to filtering. Every component placed between the processor and the outside world adds capacitance and loss that can degrade qubit coherence, so embedding the function in the board removes a discrete component and shortens the signal path. As qubit counts grow, that economy becomes decisive, and it explains why a printed circuit is now part of the microwave design of a quantum processor rather than an accessory to it.
Reading the Result Correctly
One experiment does not establish a manufacturing standard, and the result should be read as a demonstration that the approach works rather than as a specification others can copy. The board geometry, the material and the filter design were developed together, and reproducing the outcome requires the same combination rather than a similar component list.
What the work does show is where the technical risk sits. It is not in the concept, which is sound, but in fabrication repeatability on low loss materials at tight tolerances, and in the assembly and testing necessary to verify behaviour at operating temperature. Suppliers entering this field should expect to invest in process characterisation first, because that is the part that cannot be bought as equipment.
Where the Board Goes Next
As qubit counts increase, the pressure to move function out of the quantum chip and into surrounding hardware will continue. Boards that multiplex, filter and route at cryogenic temperatures allow the processor to be devoted to computation, which is the most expensive real estate in the system.
For design teams this suggests involving the board manufacturer early, treating radio frequency performance and mechanical behaviour as one problem, and specifying materials that are understood at the operating temperature. It is a demanding combination, and it is the reason a disciplined manufacturing process matters more here than a long list of capabilities.
