Observation of Genuine Three-Photon Interference

Multiparticle quantum interference is critical for our understanding and exploitation of quantum information, and for fundamental tests of quantum mechanics. A remarkable example of multi-partite correlations is exhibited by the Greenberger-Horne-Zeilinger (GHZ) state. In a GHZ state, three particles are correlated while no pairwise correlation is found. The manifestation of these strong correlations in an interferometric setting has been studied theoretically since 1990 but no three-photon GHZ interferometer has been realized experimentally. Here we demonstrate three-photon interference that does not originate from two-photon or single photon interference. We observe phase-dependent variation of three-photon coincidences with $(92.7\pm 4.6)\%$ visibility in a generalized Franson interferometer using energy-time entangled photon triplets. The demonstration of these strong correlations in an interferometric setting provides new avenues for multiphoton interferometry, fundamental tests of quantum mechanics, and quantum information applications in higher dimensions.

Figure 1

Three-photon Franson interferometer. (a) Each of three energy-time entangled photons (at wavelengths 842, 1530, and 1570 nm) travels through an unbalanced interferometer with a path difference $\tau$ between the short ($S$) and long ($L$) paths. (b) The measured arrival time difference histogram with a bin size of 0.78 ns and peak separation of $\tau =3.7\mathrm{ns}$ displays seven narrow peaks corresponding to the eight possible path combinations $S_1{S}_2{S}_3$, $L_1{S}_2{S}_3$, $S_1{L}_2{S}_3$, $S_1{S}_2{L}_3$, $L_1{L}_2{S}_3$, $L_1{S}_2{L}_3$, $S_1{L}_2{L}_3$, and $L_1{L}_2{L}_3$. When all three photons take either the short or the long path the relative arrival time is the same, so the $S_1{S}_2{S}_3$, and $L_1{L}_2{L}_3$ events overlap, forming the central peak. This overlap is a coherent superposition, leading to a three-photon coincidence rate that depends on the phases ${\phi }_n$ ($n=1$, 2, 3).

Figure 2

Experimental setup for the observation of genuine three-photon interference. A continuous-wave grating-stabilized laser diode (404 nm, 43 mW, $>25\mathrm{m}$ coherence length) pumps a 25 mm periodically poled potassium titanyl phosphate (PPKTP) crystal to generate pairs of $842/776\mathrm{nm}$ photons in type-II down-conversion, which are split at a polarizing beam splitter (PBS). The 776 nm photons pump a periodically poled lithium niobate (PPLN) waveguide to generate $1530/1570\mathrm{nm}$ photon pairs in type-0 down-conversion. These infrared photons are split in free-space by a dichroic mirror (DM) before entering the three-photon Franson interferometer, which is realized as three spatial modes of a single interferometer with a path difference $\tau =3.7\mathrm{ns}$. Photon phase control is achieved with motorized glass plates. At the two output ports $A$ and $B$, the 842 and $1530/1570\mathrm{nm}$ photons are detected with free-running silicon avalanche photodiodes (Si-APD) and superconducting nanowire single photon detectors (SNSPD), respectively, and their arrival time is registered with a time tagger system. All fibers (yellow) are single-mode fibers at respective wavelengths. A few pump photons are picked off and sent through another interferometer path ($S$—not drawn) for interferometer stabilization (see Sec. I of Ref. [18]).

Figure 3

Three-photon coincidences, two-photon coincidences, and single photon counts in our three-photon Franson interferometer. The measured three-photon coincidences (a) show clear signature of interference with an average visibility of $(92.7\pm 4.6)\%$ without background subtraction. The error bars are approximated by Poissonian count errors. No systematic modulation is visible in the measured (b) two-photon coincidences that can lead to a triplet in the histogram Fig. 1 and 1 single detection rates. The letters in the legend of the twofold coincidences indicate the set of detector combinations. For example, $1530/1570$ $AA$ is the sum of $1530/1570$ coincidences in detector combinations $A_2{A}_3$ and $B_2{B}_3$. The shown single detection rates for the $1530/1570\mathrm{nm}$ photons are dominated by dark counts of the SNSPDs, while the 842 nm dark counts (Si-APDs, $\sim 2400\mathrm{per}\sec$) are negligible.

Figure 4

The phase scan for (a) 1530 and (b) 842 nm photons of photon triplets provide further evidence for genuine three-photon interference, yielding average interference visibilities of $(84.6\pm 6.3)\%$ and $(84.6\pm 4.1)\%$, respectively.