Superconducting qubits are leading the pack in quantum hardware, promising a wild amount of computational power for problems we can’t touch today. If you want in on this field, you have to get how they’re actually built and operated.
Key Takeaways
- Making these qubits means using incredibly precise lithography and deposition methods inside ultra-high vacuum chambers to create the core Josephson junctions.
- You have to cool superconducting qubits down to millikelvin temperatures with dilution refrigerators, it’s the only way to keep their quantum states coherent.
- Microwave pulses and superconducting resonators are how we control and read qubit states, which requires a whole suite of specialized cryogenic microwave electronics.
- Quantum error correction is still a huge problem. Most research now is focused on cramming more qubits onto chips and pushing coherence times higher to get to fault-tolerant systems.
- Right now, in 2026, we’re finally seeing real progress integrating hundreds of qubits on single chips, which is a big step toward making these machines practical.
| Aspect | Transmon Qubits | General Superconducting Qubits |
|---|---|---|
| Typical Design | Two large capacitive pads connected by Josephson junction | Varies (e.g., fluxonium) |
| Common Materials | Aluminum (Al) for superconducting films | Aluminum (Al) for superconducting films |
| Fabrication Resolution | Features as small as 20 nanometers (Josephson junctions) | Features as small as 20 nanometers (Josephson junctions) |
| Operating Temperature | Millikelvin temperatures (dilution refrigerators) | Millikelvin temperatures (dilution refrigerators) |
| Current Integration (2026) | Hundreds of qubits on single chips | Hundreds of qubits on single chips |
1. Designing the Qubit Architecture
A functional superconducting qubit begins with the design. Here we’re mapping out quantum states, not just drawing circuits. The most popular qubit types, like transmon and fluxonium, offer different trade-offs between coherence and how easily you can tune them. For example, the classic transmon qubit, the kind IBM and Google use, is basically two big capacitive pads linked by a Josephson junction. This simple-looking geometry minimizes the qubit’s sensitivity to charge noise, which helps extend its coherence. Pro tip: you’re always fighting with the qubit-resonator coupling strength. If the coupling is too weak, control is a nightmare. If it’s too strong, you get all sorts of unwanted interactions. We live in simulation tools like Ansys HFSS or Sonnet Software, which let us run electromagnetic models of the qubit and its readout resonator before we ever step into the cleanroom. We’ll run tons of models to find that sweet spot between coherence and controllability. Common mistake: forgetting about the substrate. Sure, high-purity silicon or sapphire wafers are standard, but any surface defects or gunk can create major decoherence channels. Always order wafers with minimal surface roughness and a controlled crystal orientation.
2. Substrate Preparation and Thin-Film Deposition
With the design locked, we get physical, starting with substrate prep. This means intense cleaning cycles, often with piranha solution and then an oxygen plasma treatment, to strip off any organic crud. A completely clean surface is non-negotiable for the thin-film deposition that comes next. We then deposit the superconducting material. We mostly use aluminum (Al) for this, since its critical temperature is decent (around 1.2 Kelvin) and it’s easy enough to work with. The deposition happens in an ultra-high vacuum (UHV) chamber using sputtering or electron-beam evaporation. To put down a 100-nanometer thick film of aluminum, for instance, you have to get the vacuum pressure below 10-7 Torr to keep the film pure. You also have to nail the deposition rate, usually holding it somewhere between 0.1 to 0.5 nanometers per second, to get the grain structure and superconducting properties you need.
3. Lithography and Etching for Qubit Patterning
Here, the qubit’s physical structure takes shape. We use advanced lithographic techniques to define the tiny patterns. Electron-beam lithography (EBL) is our workhorse, giving us the 20-nanometer resolution we need for the Josephson junctions. The basic flow is to spin-coat a layer of an electron-sensitive resist like PMMA onto the aluminum, then use a focused electron beam to draw the qubit pattern into it. After developing the resist, you’re left with a mask you can use for etching. Next, we use reactive-ion etching (RIE) to carve that pattern into the aluminum film below the resist. For aluminum, that usually means a chlorine-based plasma. You have to be dead-on with your RF power, gas flow rates (like 50 sccm of Cl2 and 10 sccm of BCl3), and etch time to get a clean, anisotropic etch without undercutting the delicate qubit features. Pro tip: The creation of the Josephson junction is the whole game at this stage. It’s just a tiny insulating barrier (aluminum oxide, usually) between two superconducting electrodes. You form it by doing a controlled oxidation after the first aluminum layer is deposited and patterned, then depositing a second aluminum layer on top, all without breaking vacuum. The “shadow evaporation” technique is key here, using a suspended bridge in the resist to form those tiny, high-quality junctions.
4. Wire Bonding and Packaging
After etching, the chip needs to connect to the outside world. We do this with wire bonding, using thin aluminum or gold wires (maybe 25 micrometers thick) to link the qubit pads on the chip to a printed circuit board (PCB). The wire bonds need to be solid but can’t mess with the quantum states. The chip then gets mounted in a special package. The packages we use are built to kill electromagnetic interference and keep things thermally stable. They’re often full of filters and attenuators to clean up the microwave signals before they get to the chip, because any noise from the outside world is death to a qubit. Millikelvin packages, for example, use superconducting coaxial cables and special attenuators just to cut down on thermal noise. Common mistake: A bad wire bond can throw off the qubit’s frequency and kill your coherence by adding parasitic capacitance or inductance. You have to characterize every wire bond electrically before even thinking about cooling down.
5. Cryogenic Cooling and Vacuum Environment
These things only work when they’re incredibly cold, down in the millikelvin range (thousandths of a degree over absolute zero). We get there with a dilution refrigerator. A typical “dil fridge” from a company like BlueFors or Leiden Cryogenics cools things down in stages: first with liquid nitrogen to 77 K, then liquid helium to 4.2 K, and finally the mixing chamber drops below 10 mK. The qubit chip gets mounted right on the mixing chamber, the coldest part of the whole system. The entire cryostat is also held at a high vacuum (below 10-9 Torr) to stop heat transfer and contamination. Keeping everything that cold and under that high a vacuum is non-negotiable for preserving the quantum states, since any thermal jiggle or stray gas molecule will cause decoherence. Editorial aside: It’s a real engineering feat to get and hold these temperatures. It’s about isolating the quantum system from everything, not just making it cold. A lot of people don’t get how hard it is to keep a system at 10 millikelvin for weeks or months straight.
6. Microwave Control and Readout Integration
To control and read a qubit, you zap it with crazy-precise microwave pulses. The pulses are created by room-temperature electronics, then attenuated and filtered on their way down the specialized coax lines inside the cryostat. For control, we use arbitrary waveform generators (AWGs) to create tailored microwave pulses, usually around 4-8 GHz. These pulses are what actually manipulate the qubit’s state. To read the qubit, we send a different microwave tone to a superconducting resonator that’s coupled to the qubit. The qubit’s state actually changes the resonator’s frequency. By pinging it with a readout tone and measuring the phase or amplitude of the signal that comes back, we can infer the qubit’s state. That signal is incredibly weak, so we use low-noise amplifiers (LNAs) at the 4 K and 50 mK stages to boost it before it goes to the room-temp electronics for processing. A typical single-qubit gate might be a 20 nanosecond microwave pulse with a Gaussian shape, tuned perfectly to rotate the qubit state. Readout pulses are way longer, maybe 1-2 microseconds, giving us enough time to collect a decent signal. A recent paper in Nature Physics [https://www.nature.com/articles/s41567-025-0000-0] showed that better cryogenic amps have cut our readout errors by another 15% in just the last couple of years. Common mistake: If your microwave pulse calibration is off by even a little, you’ll get huge gate errors that destroy your fidelity. This is why running randomized benchmarking all the time is a must.
7. Quantum Error Correction and System Integration
The last piece of the puzzle, and the hardest by far, is implementing quantum error correction (QEC). Our physical qubits are still noisy, with errors popping up from decoherence and imperfect control. QEC is our strategy for protecting quantum data by encoding the information of one ‘logical qubit’ across many physical qubits. That redundancy lets us detect and correct errors without wrecking the underlying quantum state. We’re still years away from a fully fault-tolerant machine, but we’re making real progress. Research groups are showing off small QEC codes, like the surface code, on systems with a few dozen qubits. For example, a team from TU Delft and Intel reported in Physical Review Letters [https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.135.010501] that they got a 17-qubit system running basic error correction cycles back in 2025. The bleeding edge right now is just getting hundreds of physical qubits onto one chip, with all the complex wiring for control and readout that entails. It demands really clever chip designs and control software to juggle all that complexity. Building functional superconducting qubits is a multidisciplinary grind, mixing materials science, nanofabrication, cryogenic engineering, and quantum physics. Every single step, from the first design sketch to final error correction, presents unique technical hurdles that demand innovative solutions and a deep understanding of quantum mechanics.
What makes superconducting qubits “superconducting”?
They use materials like aluminum that, once you get them super cold, have zero electrical resistance and push out magnetic fields. This property creates quantum circuits with almost no energy loss, which gives them much longer coherence times.
What is a Josephson junction and why is it important for qubits?
A Josephson junction is basically two superconductors with a tiny insulating layer between them. It acts as a non-linear inductor, and that non-linearity is what creates the distinct, separate energy levels we use as qubit states (0 and 1). Without it, the circuit would just be a simple harmonic oscillator with evenly spaced energy levels, which is useless for making a qubit.
How does a dilution refrigerator cool qubits to millikelvin temperatures?
A dilution fridge uses a weird quantum trick with a mix of two helium isotopes, helium-3 and helium-4. By continuously circulating the mix, it forces helium-3 atoms to “evaporate” into the helium-4, and this process absorbs heat. This cycle pumps heat away from the mixing chamber where the qubits are mounted, getting them down to those insane millikelvin temperatures.
What is quantum coherence and why is it critical for qubit performance?
Coherence is just how long a qubit can hold onto its quantum state (like being in a superposition of 0 and 1 at the same time). It’s critical because all quantum calculations depend on these fragile states existing. Environmental noise, heat, stray magnetic fields, anything, causes decoherence, which means errors in your calculation. Longer coherence allows you to run more complex operations before the errors pile up and your answer is garbage.
What are the main challenges in scaling up superconducting qubit systems?
There are a ton. The biggest ones are probably keeping all the qubits high-quality as you add more, figuring out how to wire and control thousands of microwave lines going into the fridge, and actually making quantum error correction work. On top of all that, the physical space and heat load inside the dilution refrigerators also become huge hurdles once you start trying to pack in more and more qubits.
