Experimental Realization of Quantum Illumination

We present the first experimental realization of the quantum illumination protocol proposed by Lloyd [Science 321, 1463 (2008)] and S. Tan et al. [Phys. Rev. Lett. 101, 253601 (2008)], achieved in a simple feasible experimental scheme based on photon-number correlations. A main achievement of our result is the demonstration of a strong robustness of the quantum protocol to noise and losses that challenges some widespread wisdom about quantum technologies.

Figure 1

Experimental setup. (a) Quantum illumination: after the beta-barium borate (BBO) crystal, where TWB are generated, one of them (the “ancilla”) is reflected toward the detection system. The correlated beam is partially detected, together with the thermal field from the Arecchi’s disk, when the object (actually a beam splitter) is present, otherwise it is lost (not shown). A low-pass filter (95% of the transmission at 710 nm) and a UV-reflecting mirror (not shown) are used to minimize the background noise while maintaining low losses. The lens, placed at the focal length from the crystal and the CCD camera, realizes the Fourier transform of the field at the output face of the crystal. (b) Classical illumination: one beam from PDC is stopped and the other one is split at a BS for generating correlated multithermal beams. The power of the pump is adjusted to obtain the same energy of the TWB. (c) Detected TWB, in the presence of the object, without thermal background. The region of interest is selected by an interference filter centered around the degeneracy wavelength (710 nm) and a bandwidth of 10 nm. After selection, the filter is removed. (d) Detected field for split thermal beams in the presence of the object, without thermal noise. (e) A typical frame used for the measurement where a strong thermal background has been added on the object branch. The color (intensity) scales on the right correspond to the number of photons per pixel.

Figure 2

Generalized Cauchy-Schwarz parameter $\epsilon$ in the case of twin beams, ${\epsilon }^{\mathrm{TW}}$, and of the correlated thermal beams, ${\epsilon }^{\mathrm{TH}}$, as a function of the average number of background photons $⟨N_b⟩$ for a number of background modes $M_b=57$ (black squares series) and $M_b=1300$ (red circles and diamonds). The lines represent the theoretical prediction at $\mu =0.075$ (the last estimated independently).

Figure 3

Covariance in the presence, ${\Delta }_{1,2}^{\mathrm{in}}$ (dark blue), or absence, ${\Delta }_{1,2}^{\mathrm{out}}$ (light green), of the target. Panels (a) and (b) refer to QI and CI, respectively, for the same number of background modes $M_b=1300$; panel (c) refers to QI with a lower number of modes, $M_b=57$. Uncertainty bars represent the effect of the background noise on the covariance estimation (obtained averaging over the $N_{\mathrm{img}}=2000$, 6000, and 4000 images in the graphs from the top to bottom, respectively). Horizontal lines are the theoretical values $⟨{\Delta }_{1,2}^{\mathrm{in}/\mathrm{out}}⟩$, while the dashed lines are the uncertainty interval evaluated theoretically as $⟨{\delta }^2{\Delta }_{1,2}^{\mathrm{in}/\mathrm{out}}⟩/\sqrt{N_{\mathrm{img}}}$.

Figure 4

Signal-to-noise ratio (SNR) versus the number of background photons $⟨N_b⟩$ normalized by the square root of the number of realization. The red (black) markers refer to $M_b=1300$ ($M_b=57$) and the solid (dashed) theoretical curve corresponds to quantum (classical) illuminating beams. The estimation of quantum mean values of Eq. (2) is obtained by performing averages of ${\Delta }_{1,2}^{\mathrm{in}/\mathrm{out}}$ over a set of $N_{\mathrm{img}}$ acquired images ($N_{\mathrm{img}}=2000$, 4000, and 6000 for twin beams at $M_b=1300$, $M_b=57$, and thermal beams at $M_b=1300$, respectively). The lowest curve of the classical protocol has not been compared with the experimental data because the SNR is so low that a very large number of images (out of the possibility of the actual setup) is required to get reliable points.

Figure 5

Error probability $P_{\mathrm{err}}$ of the target detection versus the total number of photons of the thermal background $N_b$ evaluated with $N_{\mathrm{img}}=10$ ($N_{\mathrm{img}}=100$ in the inset). The black squares and red circles are the data for QI with $M_b=57$ and $M_b=1300$, respectively, while red diamonds refer to the data for the CI with $M_b=1300$. The curves are the corresponding theoretical predictions.