Amplification approaching the limits of physics
Quantum bits (qubits) – the basic unit of information in quantum computing – are delicate systems. In superconducting quantum computers, the energy difference between a zero and one state of a qubit is typically less than one thousandth of the kinetic energy of an average air molecule around us. Therefore, minute electrical signals must be used to probe the qubits to get a faithful reading of their state, without the readout signal itself causing transitions between the qubit states.
Our amplifiers are placed close to the QPU to increase the amplitude of these extremely weak signals encoding the qubit states. The uncertainty principle mandates that any amplifier like this must add a certain minimum quantity of noise to the amplified signal. In our factory testing, we verify that each individual amplifier we ship gets close to this so called standard quantum limit (SQL).
In the readout solution as a whole, this first amplification step is crucial because it effectively determines the signal-to-noise ratio and ultimately the fidelity of the qubit readout. The better the first amplifier is, the higher the likelihood of correctly determining whether the qubit was in the zero or one state. Until recently, the best amplifiers were available only as one-of-a-kind prototypes. We are delighted to announce that we can produce and test these state-of-the-art amplifiers in volumes sufficient for even the most aggressive scale-up plans.
AI-TWPA-C

Our flagship product is a C-band (4-8 GHz) travelling wave parametric amplifier that uses three-wave mixing and flux-tunability to achieve wideband near-quantum-limited amplification. In comparison to simpler four-wave mixing amplifiers, AI-TWPA-C has the following fundamental advantages:
- The pump frequency is far above the signal band, making it easier to block its way toward the sensitive device under test (qubits).
- No need for dispersion-engineering-related resonances within or close to the signal band.
- Avoiding the four-wave mixing Kerr nonlinearity increases the dynamic range.
- The flux-tunability adds a third dimension to the control parameter space, in addition to pump frequency and power. Thanks to it, a single AI-TWPA-C amplifier has multiple good operation points that the user may freely switch between in situ, depending on the application and system requirements.
On a more practical note:
- With an on-chip integrated flux line* and industry’s smallest form factor, you can pack them very densely, despite the flux-degree of freedom.
- Niobium-based junctions ensure reliable operation essentially indefinitely over many cooldown-warmup cycles (no aging). Tune your amplifier parameters once after installation, and you’re done.
- Mature and scalable fabrication allows us to meet even the most optimistic scenarios for the growth of channel count.
- Automated factory testing allows us to select devices with operation points tailored to your specific needs.
* Our patented on-chip flux line generates extremely uniform flux bias and negligible off-chip fringing fields, in contrast to the more conventional solution of using an external coil for providing magnetic flux bias.
Qubit readout using AI-TWPA-C

Here, we take a look at how AI-TWPA-C is used to amplify a typical qubit readout signal. Besides the TWPA, the readout chain requires several auxiliary passive components such as isolators and a diplexer, as indicated in the diagram. We have demonstrated good performance with components from multiple providers and can provide recommendations best suited to your needs and existing setup.
On a high level, the dispersive readout scheme, almost universally used to read out superconducting qubits nowadays, relies on measuring a small qubit-state-dependent shift in the frequency of a so-called readout resonator coupled to the qubit. This elegant scheme can provide near-ideal quantum-non-demolition measurements since the interaction of the qubit with the external world is restricted to well-controlled interactions with photons in the relatively-far-detuned readout resonator, and all resonant interactions at the qubit frequency remain heavily filtered throughout the entire readout process.
An important practical benefit of the dispersive readout scheme is its natural compatibility with frequency multiplexing. That is, if the amplification chain is broadband enough, multiple pairs of qubits and readout resonators can be coupled to the same physical readout line and readout simultaneously, as long as the frequencies of the readout resonators are chosen to be different. Here, besides the noise performance, the bandwidth and the dynamic range of the TWPA are essential parameters in determining how many qubits can be frequency-multiplexed into one readout line. Here, AI-TWPA-C provides industry-leading performance, with the bandwidth and input-referred 1dB compression point typically exceeding 2 GHz and -90 dBm, respectively.
Here is a more practical and slightly simplified microwave-engineer perspective of the readout scheme depicted in the diagram:
- When we want to start the qubit state measurement, we turn on the probe signal (Probe input). For the sake of simplicity, let’s choose the carrier frequency to be equal to the readout resonator frequency corresponding to the qubit being in its ground state (the zero state). This frequency is typically between 5 and 8 GHz.
- Within the qubit device (dashed box), the probe propagates in a transmission line that is weakly coupled to the relatively-high-Q readout resonator (the meander line representing a CPW resonator). As you may recall, a high-Q resonator coupled to a transmission line forms a bandstop filter. Thus, at low powers where the response is linear, the probe signal is reflected if the qubit is in its ground state, whereas it will pass through if the qubit is in its excited state, since the excitation of the qubit shifts the bandstop away from the probe frequency by a few MHz.
- Next the probe signal passes to the readout chain. Isolators (arrow symbol) on one hand protect the sensitive qubit device from back-propagating noise, and on the other hand improve the impedance matching and directivity of the TWPA.
- The diplexer (low-pass+high-pass filter symbol) allows the probe signal carrying the qubit state information to pass right through to AI-TWPA-C, while simultaneously providing a third port through which the much-higher-frequency AI-TWPA-C pump tone can be provided.
- In AI-TWPA-C, the probe signal and the pump travel through a meander of thousands of Josephson junctions that provide low-loss nonlinearity that transfers energy from the pump tone to the probe signal, thus providing near-quantum-limited amplification. The two dc lines provide magnetic flux biasing used to tune the characteristics of AI-TWPA-C, using an on-chip flux bias line.
- At the output of AI-TWPA-C, the amplified probe signal, the pump and the so called idler emerge. The pump is then filtered away by a low-pass filter to avoid saturation of the high-electron mobility transistor (HEMT) at the 3 K stage, and of the components that follow it.
- The signal leaves the cryostat and is further amplified, down converted, filtered, and digitized by suitable instrumentation (not shown).
- After some tens or hundreds of nanoseconds of averaging the amplified probe output, the signal-to-noise ratio will be high enough that we can confidently tell whether the probe is there or not, and thus whether the qubit is in its ground state or not.
- We can then turn the probe input off.