Josephson Parametric Amplifier

A superconducting resonator whose nonlinear spring constant is written by quantum mechanics, amplifying microwave whispers so faint they carry the noise of the vacuum itself.

A Josephson parametric amplifier receives a microwave signal — typically the 4–8 GHz readout pulse reflected from a superconducting qubit resonator — and amplifies it by 15–30 dB before passing it to a room-temperature semiconductor amplifier chain. Without a JPA, the added noise of conventional cryogenic amplifiers (HEMT low-noise amplifiers) would obscure the single-photon-level qubit state signal. With a JPA at the first stage, operating at 20 mK inside a dilution refrigerator, the total system noise approaches the standard quantum limit: a phase-preserving amplifier adds at least ½ photon of noise at signal frequency (the Caves 1982 limit), and a JPA achieves near this bound in practice. That noise reduction enables single-shot qubit readout fidelity above 99% in state-of-the-art superconducting quantum processors.

The JPA is built from a Josephson junction — two superconductors separated by a thin insulating barrier — embedded in an LC microwave resonator (typically a quarter-wave or lumped-element coplanar waveguide resonator). The Josephson junction's inductance is nonlinear: it depends on the phase difference φ across the junction as L_J = Φ₀/(2π I_c cosφ), where Φ₀ = h/2e ≈ 2.07×10⁻¹⁵ Wb is the magnetic flux quantum and I_c is the critical current. This nonlinear inductance makes the resonator frequency sensitive to the signal amplitude — the parametric amplification mechanism. In operation, a pump tone at twice the resonator frequency (degenerate mode) or at the sum of signal and idler frequencies (non-degenerate mode) continuously modulates the junction inductance, transferring energy from the pump to the signal. In degenerate (phase-sensitive) operation, the amplifier can squeeze noise below the half-photon limit on one quadrature at the expense of increased noise on the conjugate quadrature. A SQUID (superconducting quantum interference device) — a loop with two junctions — replaces the single junction in most designs because an external magnetic flux can tune its effective I_c, allowing the operating frequency to be adjusted without changing hardware. A ferrite circulator is placed before the JPA to separate the incoming signal from the amplified reflection, preventing pump leakage and back-action on the qubit.

In plain terms

Imagine you're trying to hear someone whisper from across a football field. You need a microphone that makes the whisper louder without adding its own hiss. The Josephson parametric amplifier is the quietest microphone physics allows: quantum mechanics says any amplifier must add at least a tiny amount of its own noise, and the JPA operates right at that absolute limit.

Also called: JPA, Josephson paramp, SQUID parametric amplifier, quantum-limited microwave amplifier.

Why it matters

Single-shot qubit readout fidelity above 99% is only achievable when the first-stage amplifier adds noise near the quantum limit. The JPA is the practical solution at millikelvin temperatures, where semiconductor amplifiers cannot operate reliably.
The Caves 1982 theorem establishes that any phase-preserving linear amplifier adds at least half a photon of noise at the signal frequency. JPA demonstrations have reached within a factor of 1.5–2× of this bound in production readout chains.
In degenerate (phase-sensitive) mode, a JPA produces squeezed microwave vacuum states — states where quantum noise is redistributed below the vacuum level on one quadrature. Squeezed microwave states are a resource for quantum communication and sensing beyond the standard quantum limit.
Every superconducting quantum processor from IBM Quantum, Google, and Rigetti that claims high-fidelity readout relies on JPA (or its close relative, the JTWPA) as the first amplification stage.

Where it lives

IBM Quantum Heron and Eagle processors — JPA readout chains in dilution refrigerators for all qubit measurement lines
Google Sycamore processor — travelling-wave parametric amplifiers (JTWPA, a distributed JPA variant) for 53+ qubit readout
HAYSTAC axion dark-matter search (Yale/Berkeley/Colorado) — JPA operating in squeezed mode to circumvent the standard quantum limit in cavity haloscope readout
ADMX (Axion Dark Matter eXperiment, University of Washington) — SQUID-based and JPA-based readout for sub-GHz dark matter searches
Quantum Circuits Inc. (QCI) modular quantum computing — JPA integrated in qubit modules for commercial cloud access

Key numbers

Operating temperature: <100 mK (typically ~20 mK) (Inside a dilution refrigerator; thermal photon occupation at 6 GHz and 20 mK is ~3×10⁻⁷, negligible compared to the quantum half-photon noise floor)
Signal frequency range: 4–8 GHz (Standard superconducting qubit readout band; JPAs can be designed for 1–12 GHz with appropriate resonator geometry)
Gain: 15–30 dB (Typical at operating point; higher gain reduces bandwidth (gain-bandwidth product is approximately fixed for a given pump power))
Instantaneous bandwidth: 10–500 MHz (Phase-preserving lumped-element JPA; travelling-wave JPAs (JTWPAs) can reach several GHz of bandwidth at 15 dB gain)
Added noise (quantum limit): ½ photon (phase-preserving) (Caves 1982 bound: n_add ≥ ½. Best JPA demonstrations achieve n_add ≈ 0.5–1.0 in practice)
Flux quantum (Φ₀): h/2e ≈ 2.068×10⁻¹⁵ Wb (Fundamental to JPA operation: sets the periodicity of SQUID inductance tuning with applied flux)

Surprising facts

The Caves 1982 quantum limit states that a phase-preserving amplifier must add at least n_add ≥ ½ photon of noise energy at the signal frequency. At 6 GHz, half a photon corresponds to an energy of 2×10⁻²⁴ J — roughly 50 million times less energy than a single photon of visible light. The JPA operates near this limit.
A Josephson tunnel junction is typically 100–200 nm on a side and 1–2 nm thick — the insulating barrier is only a few atomic layers. Yet the phase of the superconducting order parameter across this atomic sliver determines the junction's inductance to a precision that allows 20 dB of gain with sub-photon noise.
In squeezed-state operation, a JPA reduces the vacuum noise on one quadrature below the vacuum level. The HAYSTAC experiment used this squeezing to search for axion dark matter with sensitivity 3.5 dB below the standard quantum limit — the first time quantum squeezing was used to exceed the SQL in a dark-matter search.
Brian Josephson predicted the effect bearing his name in 1962 as a 22-year-old PhD student at Cambridge. He was awarded the Nobel Prize in 1973 — the prediction-to-Nobel interval is one of the shortest in physics history.

Open questions

Quantum noise cannot be amplified away: The Caves 1982 theorem is not a practical limitation waiting to be engineered around — it is a direct consequence of the commutation relations of quantum fields. Any phase-preserving linear amplifier that adds less than half a photon of noise would allow the construction of a device violating the Heisenberg uncertainty principle for the electromagnetic field. The JPA is already within experimental uncertainty of this limit in laboratory demonstrations, meaning further noise reduction is not possible for phase-preserving amplification. The only routes to lower effective noise are phase-sensitive operation (which requires knowing the quadrature of the signal in advance) or quantum error correction at the qubit level.

A short history

Brian Josephson predicted the DC and AC Josephson effects in 1962 during his PhD at Cambridge, work that earned him the 1973 Nobel Prize in Physics. Early parametric amplification concepts using Josephson devices appeared in the late 1970s with work by Wahlsten, Rudner, and Cleason. The canonical theoretical framework for Josephson parametric amplification was laid out by Yurke, Corruccini, Kaminsky, and colleagues at Bell Labs in their 1989 Physical Review A paper, which derived the gain, noise, and squeezing relations still used today. The practical era began in 2007 when Miguel Castellanos-Beltran and Konrad Lehnert at NIST Boulder demonstrated a JPA that produced microwave squeezing, with 10 dB of noise reduction below the vacuum level in one quadrature — a landmark result. From 2010 onward, JPAs were integrated into superconducting qubit readout chains at IBM, Google (then Santa Barbara), Yale, and ETH Zürich, enabling the high-fidelity qubit measurement underpinning fault-tolerant quantum computing research. The travelling-wave version (JTWPA, 2015, Macklin et al., Science) extended JPA bandwidth from tens of megahertz to several gigahertz, enabling simultaneous readout of many qubits on a single amplifier.

Sources

Caves C.M. — 'Quantum limits on noise in linear amplifiers', Phys. Rev. D 26(8), 1817–1839 (1982). DOI: 10.1103/PhysRevD.26.1817 (https://doi.org/10.1103/PhysRevD.26.1817)
Yurke B. et al. — 'Observation of 4.2-K equilibrium-noise squeezing via a Josephson-parametric amplifier', Phys. Rev. Lett. 60, 764 (1988); and Yurke et al., Phys. Rev. A 39, 2519 (1989). DOI: 10.1103/PhysRevA.39.2519 (https://doi.org/10.1103/PhysRevA.39.2519)
Castellanos-Beltran M.A. & Lehnert K.W. — 'Widely tunable parametric amplifier based on a superconducting quantum interference device array resonator', Appl. Phys. Lett. 91, 083509 (2007). DOI: 10.1063/1.2773988 (https://doi.org/10.1063/1.2773988)
Roy A. & Devoret M. — 'Introduction to parametric amplification of quantum signals with Josephson circuits', Comptes Rendus Physique 17(7), 740–755 (2016). DOI: 10.1016/j.crhy.2016.07.012 (https://doi.org/10.1016/j.crhy.2016.07.012)
Macklin C. et al. — 'A near–quantum-limited Josephson traveling-wave parametric amplifier', Science 350(6258), 307–310 (2015). DOI: 10.1126/science.aaa8525 (https://doi.org/10.1126/science.aaa8525)