letters to nature

.............................................................. In order to confirm this for our samples, we have measured

polarization-resolved transmission spectra of one of the hole arrays
Plasmon-assisted transmission of for various angles of incidence v inc of plane-wave radiation (Fig. 1b
entangled photons and c). Angle tuning is expected to shift the various resonances in
the transmission spectrum in different ways. As we have theoreti-
cally associated the peak at 809 nm with a (^1,^1) SP, we have
E. Altewischer, M. P. van Exter & J. P. Woerdman
varied the angle of incidence by tilting the samples along the
Leiden University, Huygens Laboratory, PO Box 9504, 2300 RA Leiden, diagonal axis of the square hole array. For incident light polarized
The Netherlands orthogonal to this tilting axis (Fig. 1b), the 809-nm peak splits and
............................................................................................................................................................................. shifts for increasing v inc; for incident light polarized along the tilting
The state of a two-particle system is said to be entangled when its axis (Fig. 1c), the 809-nm peak remains at approximately the same
quantum-mechanical wavefunction cannot be factorized into spectral position. These results confirm the association of the
two single-particle wavefunctions. This leads to one of the 809-nm peak with the (^1,^1) SP.
strongest counter-intuitive features of quantum mechanics, We generate polarization-entangled twin photons at a wavelength
namely non-locality1,2. Experimental realization of quantum of 813 nm with the standard method of type II spontaneous
entanglement is relatively easy for photons; a starting photon parametric down-conversion11,12 depicted in Fig. 2. The twin
can spontaneously split into a pair of entangled photons inside a
nonlinear crystal. Here we investigate the effects of nanostruc-
tured metal optical elements3 on the properties of entangled
photons. To this end, we place optically thick metal films
perforated with a periodic array of subwavelength holes in the
paths of the two entangled photons. Such arrays convert photons
into surface-plasmon wavesoptically excited compressive
charge density waveswhich tunnel through the holes before
reradiating as photons at the far side47. We address the question
of whether the entanglement survives such a conversion process.
Our coincidence counting measurements show that it does, so
demonstrating that the surface plasmons have a true quantum
nature. Focusing one of the photon beams on its array reduces the
quality of the entanglement. The propagation of the surface
plasmons makes the array effectively act as a which way detector.
Light can potentially couple to surface plasmons (SPs) if the
surface in which the SPs reside shows a periodic structure to satisfy
conservation of energy and momentum. The samples that we have
studied are two metal hole arrays. Each array is 1 mm 1 mm, and
comprises a 200-nm-thick gold layer (on glass) perforated with a
square grid of 200-nm-diameter holes spaced with a 700-nm lattice
constant; Fig. 1a inset shows a typical scanning electron microscope
picture. This figure also shows transmission spectra of the two
arrays, measured with a spectrometer using normally incident white
light. We can clearly see the resonances due to excitation of SPs on
either of the metaldielectric boundaries. At these resonances, the
measured transmission can be orders of magnitude larger than the
value obtained from classical diffraction theory for subwavelength
holes4,8. In a simple picture, the surprisingly large transmission is
due to the coupling of a photon to an SP on one side of the metal,
subsequent tunnelling of the SP through the holes to establish an SP
at the other side, and final reradiation into a photon5. Other
prominent features in the spectra are the transmission minima
associated with Wood anomalies6. The theoretical description of the Figure 1 Wavelength-dependent transmission of the arrays used in the experiment. The
full transmission spectrum is incomplete at present, but the role of dashed vertical line indicates the wavelength of 813 nm used in the entanglement
the SP is well established57. The resonance used in our experiment is experiment. a, The transmission of the two arrays. Inset, scanning electron microscope
centred at 809 nm, and has a width of ,40 nm; a calculation based picture of part of a typical hole array. Scale bar, 2 mm. The hole arrays were produced at
on the geometry of the array and the optical constants of gold and the Delft Institute of Micro-Electronics and Sub-micron Technology (DIMES) by first
glass shows that it is associated with the (^1,^1) SP mode on the defining, with electron beam lithography, arrays of dielectric pillars on a 0.5-mm-thick
glassmetal interface. The label (^1,^1) denotes any of the four glass substrate, subsequently evaporating the gold layer onto a 2-nm titanium bonding
diagonal directions that are frequency degenerate for excitation layer, and finally etching away the pillars to leave the array of air holes. b, c, Wavelength-
under normal incidence. Peak transmissions of the two arrays at dependent transmission of one of the hole arrays for various angles of incidence at a
813 nm are typically 3% (dashed curve) and 5% (solid curve); these polarization orthogonal to the tilting axis (b) and along the tilting axis (c). In both
values are much larger than the value of 0.55% obtained from graphs the lowest curve is measured with the probe beam at normal incidence, while
classical theory8. The difference in transmission between the two consecutive curves are measured at increasing angles of incidence (one-degree steps) by
nominally identical hole arrays is ascribed to production subsequent tilting of the square hole array around a diagonal. These curves have been
imperfections. plotted with subsequent vertical offsets of 2% on the transmission scale. The resonance
Because an SP is a mainly longitudinal, compressive electron around 813 nm (dashed line) shows a complicated splitting for a polarization orthogonal
density wave, its propagation direction depends on the polarization to the tilting axis (b), whereas it is approximately stationary for a polarization along this
axis of the incident light, following a certain dispersion relation9,10. axis (c).

304 2002 Nature Publishing Group NATURE | VOL 418 | 18 JULY 2002 | www.nature.com/nature
 letters to nature
photons travel along two beam lines defined by pinholes, pass removed in our set-up by the compensating crystals C, which act as
through polarizers P1 and P2, and are detected by single-photon quantum erasers.
counters. The photodetectors act as bucket detectorsthat is, they Placement of the two hole arrays in the two beam lines leads
impose no further transverse mode selection (this is only done by naturally to a marked reduction of single and coincidence counts.
pinholes). To provide the two-photon coincidence rate, the signals Coincidence count rates are typically 55 s21 at the optimum setting
from the two detectors are combined electronically in a coincidence of the detecting polarizers, which is consistent with the trans-
circuit with a time window of 2 ns. missions of the arrays given above 3% 5% 3:2 104 s21 <
In a simplified picture, the generated polarization-entangled state 50 s21 : We again measured the purity of the entangled state, and
is found that the visibilities were now V 08 97.1% and
1 V 458 97.2%, respectively. In Fig. 3 the corresponding fourth-
jWl pffiffi jH1 V2 l eiv jV1 H2 l 1 order quantum interference fringes are shown for polarizer P2 fixed
2
at 08 and 458, and P1 varying in steps of 108. These measurements
where the state jH1V2l represents the simultaneous emission and show convincingly that the quantum entanglement almost com-
propagation of an H-polarized photon in beam 1 and a V-polarized pletely survives the transition from photon to SP and back. As a
photon in beam 2. The H- and V-directions are defined by the further confirmation, we performed a measurement of the so-called
orthogonal birefringent axes of the BBO (b barium borate) crystal S-parameter, as described in ref. 11, on a singlet Bell state (v p in
generating the twin photons, and all spatial information is equation (1)). This experiment, which took 16 runs of 100 seconds
implicitly contained in the beam labelling. By tilting one of the each, gave a value of jSj 2:71 ^ 0:02; which is a violation of the
compensating crystals (C in Fig. 2), the quantum phase v can be set. classical inequality jSj # 2; ref. 13) by 35 standard deviations.
In the absence of the hole arrays, our set-up produces typically Further experiments were done on a set-up with only a single
3.2 104 coincidence counts per second, which is ,25% of the array in one of the beams. The results for this case look very similar
single count rate. A measure for the purity of the quantum to those for the experiment with two arrays; again the entanglement
entangled state is the so-called visibility of the biphoton fringe11,12. was practically unaffected (see Table 1). This is to be expected, as the
This visibility was typically V 08 99.3% and V 458 97.0% at two-photon wavefunction of equation (1) is perturbed by changes
polarization angles of 08 and 458, respectively (see Table 1). The in either of its single-photon components; in principle, a single
high value at 458 shows that the natural preference of the BBO array could have removed all entanglement. The difference between
crystal for its birefringent axes (08 and 908) was almost completely the two single-array experiments (Table 1) is due to imperfections in
array 2, which are also observable in its (single-photon) polariz-
ation-dependent transmission. As the measurements using only
array 2 gave results very similar to the situation with both arrays in
place, these imperfections must have caused the somewhat limited
visibilities in the two-array experiment.
The most intriguing results of our single-array experiment are
obtained when we focus one of the beam lines onto its hole array,
using a confocal telescope (close to lens L) of two f 15 mm lenses
symmetrically positioned around the array, as shown inside the
large dotted rectangle in Fig. 2. Under these conditions, we observe a
notable reduction of the degree of quantum entanglement: when the
intra-telescope focus has a numerical aperture of 0.13, we observe
visibilities of V 08 73% and V 458 87% (Table 1).
The observed reduction in visibility on focusing can be explained
as a consequence of the non-local relation between the electronic
excitation in the metal film and the incident optical field; SPs are not
at all local, but instead propagate along the dielectricmetal inter-
face at nearly the speed of light over distances of many optical
wavelengths3,9,10. As a result of this propagation, the near-field
distribution of the photons that are reradiated at the back of the
Figure 2 Experimental set-up. A 240-mW continuous-wave krypton-ion laser beam at a hole array differs from the spatial profile of the polarization-
wavelength of 406.7 nm is directed onto a 1-mm-thick BBO nonlinear crystal, where the isotropic incident photons. Because we use the (^1,^1) SP
beam diameter is ,0.50 mm (full width at 1/e2 points). Inside the nonlinear crystal, a resonance (at 809 nm) we expectfor unpolarized incident light
small fraction of the pump photons is down-converted into twin photons at the doubled a near-field pattern consisting of two orthogonal ellipses at the
wavelength (813 nm); these are emitted along two intersecting cones. Polarization- back of the hole array (Fig. 2 inset). The unpolarized overlap region
entangled photon pairs are selected with pinholes at the crossings of these cones; the size of these ellipses correspond to the focused incident light, whereas
of the pinholes (far-field diameter 5 mrad) was chosen as a compromise between high the polarized protuberances introduce the possibility of dis-
yield and good quantum entanglement. Lenses L of 40 cm focal length produce a one-to- tinguishing the polarization of the photons on the basis of their
one intermediate image of the pumped area, which is used in some experiments to
accommodate the hole arrays A1 and A2. After passing through polarizers P1 and P2, the
entangled twin photons are focused through interference filters IF (10-nm bandwidth
centred at 813 nm) onto single-photon counting modules (Perkin Elmer SPCM-AQR-14).
Table 1 Biphoton fringe visibilities
Beam walkoff is compensated by the standard quantum eraser comprising a half-wave
Experiment R (s21) V 08 (%) V 458 (%)
plate HWP at 458 and compensating crystals C with a thickness equal to half of that of the .............................................................................................................................................................................
generating crystal11,12. The dotted objects are present only in some experiments; they No arrays 32 103 99.3 97.0
show the hole arrays A1 and A2 at the image positions created by lenses L, or, Both arrays 55 97.1 97.2
Only array 1 1.6 103 99.4 97.1
alternatively, in the focus of the confocal telescope TEL (15 mm focal length lenses). Inset, Only array 2 1.0 103 97.5 96.8
schematic picture of the near field at the back of array A1 when this is positioned inside Array 1, focussed 1.1 103 73 87
.............................................................................................................................................................................
the telescope. The arrows indicate the polarization direction of the optical electric field; the R, measured coincidence count rate; V 08 and V 458, measured visibility for one of the polarizers
centre region is unpolarized. fixed at 08 and 458, respectively.

NATURE | VOL 418 | 18 JULY 2002 | www.nature.com/nature 2002 Nature Publishing Group 305
letters to nature

Figure 3 Biphoton fringes. These fringes correspond to fourth-order quantum interference, and were measured with the two hole arrays in place, for P2 fixed at 08 (solid curve) and 458
(dashed curve), and P1 varying in steps of 108.

spatial near-field coordinates. This will then automatically remove instead waveguide effects. Our arrays are apparently thin enough
part of the polarization entanglement, just as any which way (thickness/period < 0.3) for waveguide effects not to play an
information will do. Note that the observed reduction in visibility important role. This conclusion is supported by experiments in
is much stronger for V 08 than for V 458, contrary to what is generally which the thickness of such a thin array has been varied6.
found without using hole arrays11,12. This observation is consistent By addressing the topic of plasmon-assisted transmission of
with the fact that we excite SPs propagating in the diagonal (458) quantum entangled photons, we have combined two fields of
directions. research, namely quantum information and nanostructured metal
The non-local nature of the electronic response is equivalent to optics. We hope that our work will stimulate other studies of the
an explicit wavevector dependence of the dielectric function transfer of entanglement to condensed-matter degrees of
(spatial dispersion; ref. 14). The latter description highlights the freedom. A
far-field aspects of a non-local dielectric response, and is related by Received 11 March; accepted 23 May 2002; doi:10.1038/nature00869.
way of a Fourier transform to the near-field picture given above.
1. Greenberger, D. M., Horne, M. A. & Zeilinger, A. Multiparticle interferometry and the superposition
A theory for the reduction of visibility due to this effect must principle. Phys. Today, 2229 (August 1993).
intrinsically be a multimode theory because, after passage through 2. Bouwmeester, D. et al. Experimental quantum teleportation. Nature 390, 575579 (1997).
the hole arrays, the spatial information within each beam should be 3. Raether, H. Surface Plasmons (Springer, Berlin, 1988).
4. Ebbesen, T. W., Lezec, H. J., Ghaemi, H. F., Thio, T. & Wolff, P. A. Extraordinary optical transmission
accounted for. Discussion of such a theory is beyond the scope of
through sub-wavelength hole arrays. Nature 391, 667669 (1998).
this Letter, but qualitative arguments are easily found. When 5. Martin-Moreno, L. et al. Theory of extraordinary optical transmission through subwavelength hole
adopting a near-field point of view, the reduction of the visibility arrays. Phys. Rev. Lett. 86, 11141117 (2001).
is expected to be significant when the propagation or coherence 6. Ghaemi, H. F., Thio, T., Grupp, D. E., Ebbesen, T. W. & Lezec, H. J. Surface plasmons enhance optical
transmission through subwavelength holes. Phys. Rev. B 58, 67796782 (1998).
length l of the SP is comparable to the transverse coherence length of 7. Grupp, D. E., Lezec, H. J., Ebbesen, T. W., Pellerin, K. M. & Thio, T. Crucial role of metal surface in
the focused incident beam. This is indeed the case for our exper- enhanced transmission through subwavelength apertures. Appl. Phys. Lett. 77, 15691571 (2000).
iment: on the basis of the spectral width of the transmission peak at 8. Bethe, H. A. Theory of diffraction by small holes. Phys. Rev. 66, 163182 (1944).
809 nm, we estimate l < 4 mm; the number of holes covered by the 9. Hecht, B., Bielefeldt, H., Novotny, L., Inouye, Y. & Pohl, D. W. Local excitation, scattering, and
interference of surface plasmons. Phys. Rev. Lett. 77, 18891892 (1996).
propagating SPs is thus about 10. In comparison, the transverse 10. Sonnichsen, C. et al. Launching surface plasmons into nanoholes in metal films. Appl. Phys. Lett. 77,
coherence length of the light is ,4 mm (set by the 5-mrad far-field 140142 (2000).
selection due to the pinholes; Fig 2). We note that the SP propa- 11. Kwiat, P. G. et al. New high-intensity source of polarization-entangled photon pairs. Phys. Rev. Lett.
75, 43374341 (1995).
gation length is an order of magnitude smaller than the value
12. Kurtsiefer, C., Oberparleiter, M. & Weinfurter, H. High efficiency entangled photon pair collection in
l < 40 mm predicted for a smooth pure gold film3; we ascribe this type II parametric fluorescence. Phys. Rev. A 64, 023802-1023802-4 (2001).
difference to radiative decay due to the hole patterning, and to extra 13. Clauser, J. F., Horne, M. A., Shimony, A. & Holt, R. A. Proposed experiment to test local hidden-
damping by the titanium bonding layer. variable theories. Phys. Rev. Lett. 23, 880884 (1969).
14. Landau, L. D., Lifshitz, E. M. & Pitaevskii, L. P. Electrodynamics of Continuous Media 2nd edn
From a general perspective, the observed conservation of quan- (Pergamon, Oxford, 1984).
tum entanglement for the conversion from photon ! surface 15. Lu, Y. J. & Ou, Z. Y. Observation of nonclassical photon statistics due to quantum interference. Phys.
plasmon ! photon is a demonstration of the true quantum nature Rev. Lett. 88, 023601-1023601-4 (2002).
16. Kimble, H. J., Dagenais, M. & Mandel, L. Photon antibunching in resonance fluorescence. Phys. Rev.
of SPs. All experiments on SPs so far have not probed their quantum
Lett. 39, 691695 (1977).
nature, but only their wave nature through their semiclassical
dispersion relation. We note that the true quantum nature of the
photon was only established in 1977, in anti-bunching exper-
Acknowledgements
iments15,16. Furthermore, a simple estimate shows that SPs are
We thank A. van Zuuk and E. van der Drift for the production of the hole arrays, and
very macroscopic, in the sense that they involve some 1010 electrons. G. Nienhuis for theoretical discussions. This work was supported by the Stichting voor
Our experiment proves that this macroscopic nature does not Fundamenteel Onderzoek der Materie (FOM), and the European Union under the
impede the quantum behaviour of SPs, because they can act as IST-ATESIT contract.
intermediates in transmitting entangled photons to yield the
expected fourth-order interference.
A theory for our experiments based only on locally induced Competing interests statement
surface currents is clearly inadequate. We stress this point because The authors declare that they have no competing financial interests.
some recent theoretical models for one-dimensional gratings in Correspondence and requests for material should be addressed to E.A.
thick metal films have questioned the role of SPs, emphasizing (e-mail: erwin@molphys.leidenuniv.nl).

306 2002 Nature Publishing Group NATURE | VOL 418 | 18 JULY 2002 | www.nature.com/nature