Quantum-Entangled Noise Radar

• Quantum-entangled noise radar is a system that employs quantum illumination by generating entangled signal-idler pairs, where the idler is stored to improve detection in noisy, lossy settings.
• It utilizes two-mode squeezed vacuum states and microwave technologies like JPAs and JTWPA to achieve a factor-of-4 error exponent improvement over classical approaches.
• Practical implementations address challenges such as idler storage loss, limited bandwidth, and spoof detection, paving the way for secure and efficient radar systems.

Quantum-entangled noise radar denotes a radar architecture in which the transmitted field is noise-like in its marginal statistics but is quantum-correlated with a retained reference, usually an idler mode stored at the receiver. In most of the technical literature this architecture is identified with quantum illumination: a signal–idler source generates entangled pairs, the signal interrogates a target region embedded in loss and thermal background, and detection is performed by testing the residual correlation between the noisy return and the retained idler. Related quantum-radar protocols also exist, including discrete-variable, Doppler-metrological, and spatial-metrological schemes, but several of those are not “noise radar” in the narrow quantum-illumination sense (Blakely et al., 2024, Mathews, 2022, Maccone et al., 2023).

1. Definition and state model

The canonical state resource is the two-mode squeezed vacuum (TMSV) generated by spontaneous parametric down-conversion or by microwave Josephson parametric devices. In the standard broadband model, time is partitioned into pulses of duration $T$, the active bandwidth is $W$, and the number of temporal–spectral modes per pulse is

$M = TW .$

For each active mode $m$, the signal–idler pair is prepared in a TMSV state

$$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$

with mean photon number per mode $N$, often $N\ll 1$. The modewise correlations include

$$\langle \hat a_{S_m}\hat a_{I_m}\rangle=\sqrt{N(N+1)},\qquad \langle \hat N_{I_m}\hat N_{S_m}\rangle = N+2N^2,$$

and the full pulse state is the tensor product over $M$ such pairs. The signal beam is transmitted, while the idler beam is stored locally; under an ideal return, the joint idler–received state is the original entangled state (Blakely et al., 2024).

In microwave implementations, the same structure is realized with superconducting parametric hardware. A Josephson parametric amplifier or Josephson Traveling Wave Parametric Amplifier can generate entangled microwave modes whose individual marginals appear thermal or noise-like, while their joint statistics exhibit the phase-sensitive correlations required by quantum illumination. The JTWPA proposal specifically targets multi-GHz entangled microwave bandwidths and reports an ultrawide bandwidth equal to $10~\mathrm{GHz}$ at X-band when pumped at $W$0, addressing the narrowband limitation of earlier JPA-based sources (Livreri et al., 2021).

2. Detection theory and quantum-illumination performance

The standard detection problem is binary hypothesis testing. Under target absence,

$W$1

where $W$2 is a thermal background mode with mean photon number $W$3. Under target presence,

$W$4

where $W$5 is the effective reflectivity and $W$6 is the transmitted signal mode. The receiver has access to the noisy return and the retained idler, and the optimum decision rule is governed by quantum hypothesis testing; for equal priors the minimum binary error is set by the Helstrom–Holevo bound

$W$7

For Gaussian quantum illumination in the high-loss, high-noise, low-brightness regime,

$W$8

the coherent-state benchmark obeys

$W$9

whereas optimal quantum illumination obeys

$M = TW .$0

This is the standard factor-of-4 improvement in the error exponent, commonly described as a $M = TW .$1 advantage over the best classical illumination with the same transmitted energy (Shapiro, 2019, Mathews, 2022).

A distinctive feature of this regime is that the return–idler state is typically separable by the time it reaches the receiver: the propagation channel is effectively entanglement-breaking. The operational advantage persists because the initial TMSV source imprints phase-sensitive signal–idler correlations that no classical transmitter with the same energy can reproduce. In the usual TMSV model, the target-present signature is the nonzero correlation

$M = TW .$2

whereas under $M = TW .$3 this quantity vanishes (Shapiro, 2019).

3. Implementations and experimental platforms

Microwave proof-of-principle systems have realized several forms of entangled noise radar. In “quantum-enhanced noise radar,” a nondegenerate JPA generates a two-mode squeezed source, both quadrature records are digitized, and target information is extracted from classical covariance processing rather than a delayed joint quantum measurement. In a direct comparison to an ideal classical noise source that saturates the classical correlation bound, the quantum source outperforms the classical source by as much as an order of magnitude in the measured covariance, even in the presence of significant added noise and loss (Chang et al., 2018).

A later microwave prototype, the quantum two-mode squeezing radar, operated entirely at microwave frequencies with entangled modes at $M = TW .$4 and $M = TW .$5. One beam propagated through $M = TW .$6 of free space, the other was measured locally, and receiver operating characteristic curves were compared to those of a classical two-mode noise radar. At a broadcast power of $M = TW .$7, the QTMS system required $M = TW .$8 times fewer integrated samples than its classical counterpart to achieve the same ROC performance (Luong et al., 2019).

Source bandwidth is a recurring systems bottleneck. JPA-based microwave quantum-radar experiments are narrowband, whereas the JTWPA architecture was proposed precisely to provide a broadband entangled-noise source. Measurement results of the developed JTWPA show an ultrawide bandwidth equal to $M = TW .$9 at X-band, and the number of independent modes scales as $m$0, so such bandwidth directly improves the usable time–bandwidth product of a quantum-illumination receiver (Livreri et al., 2021).

System-level range modeling for QTMS radar has been developed by treating correlation detection as the primary observable. In that framework, a noise radar can be interpreted as a conventional radar with a reduced threshold signal-to-noise ratio, because target detection is based on signal–idler correlation rather than received power alone. Under a feasible JTWPA-based parameter set, the analysis gives maximum detection ranges up to the order of $m$1, and distinguishes “early alarm” and “track” QTMS radars by their chosen false-alarm rates (Allahverdi et al., 2024).

A distinct discrete-variable microwave architecture has also been proposed with a quantum dot–based entangled photon generator, a transmission module, a delay line, a cryogenically cooled SNSPD-based detection module, and a signal processing unit that performs quantum state comparison. That design uses Bell-type signal–idler pairs rather than broadband TMSV noise, so it is adjacent to entangled noise radar rather than identical to the usual continuous-variable quantum-illumination model (Gautam et al., 25 Jan 2026).

4. Parameter-estimation variants and related protocols

Not all entangled radar schemes are target-detection protocols. A quantum Doppler radar based on multimode spontaneous parametric downconversion studies the estimation of the Doppler parameter $m$2 in a thermal microwave environment with loss. Using the quantum Fisher information as the metric, and matching signal energy and pulse duration between quantum and classical probes, it finds a $m$3 advantage in the regime of small signal photon number, high thermal noise, and low transmissivity. This extends the quantum-illumination logic from binary target detection to radial-velocity estimation (Wei et al., 2024).

Another line of work studies entangled Gaussian beams for time-of-flight ranging. In that protocol the beam is composed of $m$4 photons entangled in frequency degrees of freedom, and the average arrival time can be estimated with uncertainty $m$5 rather than $m$6, yielding a $m$7 enhancement over an unentangled radar. The same source explicitly distinguishes this protocol from quantum-entangled noise radar or quantum illumination in the usual sense, because it does not use a retained idler and does not address bright thermal background as the core operating condition (Maccone et al., 2023).

A related spatial-metrology protocol localizes a non-cooperative point target in three dimensions using $m$8 entangled photons and all of their spatial degrees of freedom. It reports uncertainties in each spatial coordinate that are $m$9 smaller than for $$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$0 independent photons. This is likewise an entangled radar protocol, but its natural taxonomy is quantum metrology rather than entangled noise radar (Maccone et al., 2019).

Receiver-side non-Gaussian filtering has also been proposed. A single-photon entangled radar protocol based on partially postselected filtering and photon catalysis shows that heralding and filtering can effectively increase the reflection or transmission ratio of absorbing materials. In Monte Carlo simulations with Gaussian white noise, the procedure yields a remarkable enhancement in the signal-to-noise ratio of imaging, albeit with an increase in mean-square error. A plausible implication is that entangled noise radar need not be limited to Gaussian joint measurements; probabilistic non-Gaussian receiver stages can also reshape usable contrast in the low-photon regime (Li et al., 2024).

5. Spoofing, security, and entanglement as a radar resource

Quantum-entangled noise radar has also been analyzed as a security primitive. In a spoofing model, an adversary intercepts the transmitted signal and performs a classical measure-and-prepare attack, but has no access to the retained idler and cannot recreate entanglement with it. Two explicit strategies have been studied: direct detection followed by number-state preparation, and heterodyne detection followed by coherent-state preparation. The radar operator then performs a binary hypothesis test between a true return and a spoof return (Blakely et al., 2024).

For the ideal lossless case, if the true return is the original TMSV state and the spoof is generated by direct detection plus number-state preparation, the fidelity between the two hypotheses is

$$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$1

For heterodyne detection plus coherent-state preparation,

$$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$2

Because $$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$3 for any $$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$4, direct detection produces the more dangerous spoof: it is closer in fidelity to the genuine return and therefore harder to detect. The same ordering persists when noise and loss are added to the channel model (Blakely et al., 2024).

The security significance is that a classical measure-and-prepare spoofer can reproduce some observable statistics but cannot reproduce the entangled joint state of idler and true return. The direct-detection spoof even matches the true photon-number correlation,

$$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$5

yet remains separable. By contrast, if the radar had used only classically correlated photon-number pairs,

$$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$6

the same direct-detection spoof would reproduce the state exactly, giving fidelity $$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$7. In this precise sense, entanglement functions as a non-copyable physical resource for spoof detection. The same source compares this to keyed classical radars and argues that the idler behaves like a “non-copyable, non-classically simulatable key” (Blakely et al., 2024).

6. Limitations, controversies, and outlook

The most persistent limitation is that the ideal quantum-illumination advantage assumes a retained idler that can be stored with very low loss until the return arrives. In realistic systems, idler storage loss can erase the advantage rapidly. For example, in the standard analysis a $$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$8 idler loss wipes out the full $$\ket{\psi}_m = \sum_{n=0}^\infty \sqrt{\frac{N^n}{(1+N)^{1+n}}}\; \ket{n}_{I_m}\ket{n}_{S_m},$$9 optimum quantum-illumination gain, and a $N$0 idler loss wipes out the $N$1 gain of suboptimal OPA-type receivers. The same literature also stresses that many microwave “quantum-correlated noise” radars based on heterodyne detection can be matched by properly designed classically correlated noise radars once the comparison is made against the correct classical benchmark (Shapiro, 2019).

A stronger critical position argues that there is no distinct quantum radar cross section separate from the classical radar cross section, and that in realistic microwave conditions classical noise radar outperforms quantum radar by large margins at equal energy and bandwidth. That analysis attributes the apparent gap to the severe $N$2 range law, the thermal photon number at microwave frequencies, and the requirement that genuine quantum advantage occurs only for $N$3, which conflicts with long-range power needs. On that reading, useful microwave quantum radar ranges are limited to very short distances under conservative assumptions (Galati et al., 2024).

Other work is less pessimistic but identifies a narrower set of engineering bottlenecks. Range modeling for QTMS radar points to source bandwidth as the critical parameter for achieving simultaneous quantum advantage and substantial radar range, while engineering surveys emphasize superconducting entanglement sources, transducers, amplification chains, cryogenic front ends, and receiver architectures as the core subsystems that must improve before practical deployment becomes plausible (Allahverdi et al., 2024, Karakoc et al., 12 Oct 2025).

Some literature has also proposed a “Quantum Radar Cross Section” or QRCS as a detection-oriented generalization of classical RCS, whereas critical work disputes that any physically distinct cross section exists. A cautious reading is that quantum-entangled noise radar does not change target scattering physics, but it can change the achievable decision performance for a fixed scattering response, provided the transmitter, idler storage, and receiver preserve enough nonclassical correlation to matter operationally (Mathews, 2022, Galati et al., 2024).

Across the literature, the stable conclusion is narrower than early popular claims but technically substantive. Quantum-entangled noise radar is best understood as entanglement-assisted correlation radar: it exploits a retained idler to test for joint statistics that classical illumination cannot reproduce at the same energy. In ideal theory this yields a factor-of-4 improvement in the target-detection error exponent; in experiments it has produced covariance and ROC gains at microwave frequencies; in recent security analyses it provides spoof-detection leverage because an intercept-and-retransmit adversary cannot reconstruct the idler–return entanglement (Shapiro, 2019, Luong et al., 2019, Blakely et al., 2024).