See discussions, stats, and author profiles for this publication at: [Link]

net/publication/350088020

Spontaneous Parametric Down-Conversion from Resonant Metasurfaces

Preprint · March 2021

CITATIONS READS
0 220

10 authors, including:

Tomás Santiago-Cruz Anna Fedotova

Max-Planck-Institut für die Physik des Lichts Lomonosov Moscow State University
23 PUBLICATIONS 448 CITATIONS 27 PUBLICATIONS 512 CITATIONS

SEE PROFILE SEE PROFILE

Vitaliy Sultanov Maximilian Weissflog

Max-Planck-Institut für die Physik des Lichts Friedrich Schiller University Jena
20 PUBLICATIONS 252 CITATIONS 33 PUBLICATIONS 218 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Vitaliy Sultanov on 02 March 2022.

The user has requested enhancement of the downloaded file.

 Spontaneous Parametric Down-Conversion from
Resonant Metasurfaces
arXiv:2103.08524v1 [[Link]] 15 Mar 2021

Tomás Santiago-Cruz,∗,†,‡,¶,# Anna Fedotova,§,# Vitaliy Sultanov,†,‡ Maximilian

A. Weissflog,§,¶ Dennis Arslan,§ Mohammadreza Younesi,§ Thomas Pertsch,§,¶,k

Isabelle Staude,§,⊥ Frank Setzpfandt,§ and Maria V. Chekhova†,‡,¶

†Max Planck Institute for the Science of Light, Staudtstraße 2, 91058 Erlangen, Germany.
‡University of Erlangen-Nürnberg, Staudtstraße 7/B2, 91058 Erlangen, Germany.
¶Max Planck School of Photonics, Albert-Einstein-Str. 6, 07745 Jena, Germany.
§Institute of Applied Physics, Abbe Center of Photonics, Friedrich Schiller University Jena,
07745 Jena, Germany.
kFraunhofer Institute for Applied Optics and Precision Engineering, 07745 Jena, Germany.
⊥Institute of Solid State Physics, Friedrich Schiller University Jena, 07743 Jena,
Germany.
#These authors contributed equally.

E-mail: [Link]@[Link]

Abstract

All-dielectric optical metasurfaces are a workhorse in nano-optics due to both their

ability to manipulate light in different degrees of freedom and their excellent perfor-

mance at light frequency conversion. Here, we demonstrate first-time generation of

photon pairs via spontaneous parametric-down conversion in lithium niobate quantum

optical metasurfaces with electric and magnetic Mie-like resonances at various wave-

lengths. By engineering the quantum optical metasurface, we tailor the photon-pair

1
 spectrum in a controlled way. Within a narrow bandwidth around the resonance, the

rate of pair production is enhanced up to two orders of magnitude compared to an

unpatterned film of the same thickness and material. These results enable flat-optics

sources of entangled photons – a new promising platform for quantum optics experi-

ments.

Keywords

Quantum optics, photon-pair generation, spontaneous parametric down-conversion, nonlin-

ear metasurfaces.

Introduction

An important current tendency is the miniaturization of photonic devices towards multifunc-

tional films with thicknesses in the nanometer range, 1 so-called metasurfaces. Metasurfaces
already found several applications in linear optics, like ultrathin lenses, 2 holograms, 3 mode
converters, 4 and filter structures. 5 Recently, they have also been established as a promising
platform for nonlinear optics. 6 The advantages of nonlinear flat-optics devices are their slim
profile, ultrafast and broadband operation, and relaxed phase matching in frequency conver-
sion. 7 The latter leads to unprecedented freedom in the choice and engineering of nonlinear
materials. Nanoscale sources with engineered nonlinearities and resonant field enhancement
provide efficiencies of nonlinear frequency conversion as high as 1% 8 and versatile nonlinear
beam shaping. 9 However, most of the previous work has been focused on the generation and
control of classical light. The next natural step are quantum optical metasurfaces (QOM) for
the generation of quantum states of light through nonlinear processes at the nanoscale. 10,11
While single-photon emitters have already been integrated into flat-optics platforms,
the efficient generation of photon pairs within such structures remains a challenge. Pair
generation has been reported using four-wave mixing in carbon nanotube films, 12 but the

2
efficiency and signal-to-noise ratio were low. Another commonly used nonlinear process to
create photon pairs is spontaneous parametric down-conversion (SPDC). In SPDC, a pump
photon of higher frequency ωp splits in two daughter photons, signal and idler, with lower
frequencies ωs and ωi , where energy conservation requires ωs +ωi = ωp . Recently, photon-pair
generation by SPDC was demonstrated in ultrathin films of lithium niobate (LN) and gallium
phosphide (GaP). 13,14 But despite the high bulk second-order susceptibilities χ(2) of these
materials, the achieved pair generation rates were modest: SPDC is based on the parametric
amplification of the vacuum field, 15 which is extremely weak. Furthermore, contrary to
classical nonlinear processes like second-harmonic generation (SHG), the SPDC rate scales
linearly with the pump power, and hence its efficiency cannot be enhanced by using a pulsed
or focused pump.
To boost the efficiency of nanoscale SPDC, nanostructures with resonances at the signal
and idler frequencies can be employed. Such structures feature an increased density of states,
thus enhancing vacuum fluctuations and enabling more efficient SPDC. This approach was
already studied theoretically 16,17 and experimentally 18 for single nanoresonators exhibiting
Mie-type resonances. Although the demonstrated results are promising, they are limited by
the small total volume of the used resonators. Nonlinear metasurfaces – two-dimensional
arrangements of such nonlinear nanoresonators – promise a better photon-pair generation
rate.
Especially all-dielectric high-χ(2) metasurfaces 19 are a favourable platform for SPDC,
thanks to their high damage threshold. Various resonance effects, such as Mie-type 20,21 or
Fano-type 22 resonances and bound states in the continuum, 23 have been shown to enhance
SHG efficiency by several orders of magnitude. 22,24 Due to the similarities between classical
parametric frequency conversion and SPDC, 25–27 comparable enhancements are expected for
the latter.
Here we observe, for the first time, SPDC from resonant QOMs, schematically shown in
Fig. 1(a). Due to the resonances, the photon pairs are emitted only into a narrow wavelength

3
Figure 1: (a) Artist’s view of SPDC from a LN metasurface: the pump is incident from
the substrate side, photon pairs are collected in reflection. Both the pump and the SPDC
photons are polarized along the LN optic axis z. (b) Scanning electron microscope (SEM)
image of a fabricated metasurface showing a periodic array of nanoresonators in the shape of
truncated pyramids. (c) Electric field distribution inside such a nanoresonator as calculated
in COMSOL Multiphysics at the electric (left) and magnetic (right) resonances. The incident
field is polarized along the LN optic axis. Arrows show the electric field direction. (d)
Experimental (solid lines, shading) and simulated (dashed lines) reflectance spectra of four
QOMs with different resonance positions, further labeled as A, B, C, D; vertical black line
marks the wavelength of degenerate SPDC. Vertical colored lines mark the positions of the
electric resonances.

range, which opens a possibility to engineer their spectrum. Furthermore, within the emission
bandwidth we observe two orders of magnitude enhancement of the pair generation rate
compared to an unstructured film of the same thickness as the metasurface, despite the fact
that nanostructuring reduces the volume of the nonlinear material and our optics do not
collect all photon pairs generated from the QOM. The experiment is run in the reflection
geometry, common for nonlinear optics and fluorescence experiments, but quite ‘unorthodox’
for SPDC.

Results and discussion

We generate photon pairs in QOMs made of LN, which is well known for its high second-
order susceptibility χ(2) . Several recent works already demonstrated efficient SHG with LN
nanoparticles 28 and metasurfaces. 29–32 LN is especially attractive for quantum nonlinear

4
optics due to its broad transparency range and relatively low fluorescence compared to
semiconductors like GaAs or GaP. In this work we take advantage of its largest nonlinear
(2)
tensor component χzzz .
Our metasurfaces are designed with fundamental magnetic and electric resonances for
signal and idler photons. At resonance, the density of states is increased, enabling enhanced
generation rates for photon pairs. 33 A resonance at the pump wavelength could further
increase the SPDC rate; however, since SPDC depends linearly on the pump power, the
same effect can be achieved by modifying the pump properties.
We have fabricated QOMs on a 680 nm thick x-cut LN-on-insulator film. The metasur-
faces consist of nanoresonators in the shape of truncated pyramids with side lengths around
700 nm, arranged with a period around 900 nm as shown in a scanning electron micrograph
in Fig. 1(b). Our QOMs support two Mie-like resonances in the near-infrared wavelength
range, 32 further called ‘electric’ and ‘magnetic’ for the reasons discussed below. Their field
distributions inside one nanoresonator are shown in Fig. 1(c), left and right panels, corre-
spondingly.
Using a custom-built white-light spectroscopy setup, we measured the reflectance of our
QOMs for light polarized along the optic axis of LN (Fig. 1(d), solid lines). Each QOM
exhibits a narrow electric resonance at shorter and a broad magnetic resonance at longer
wavelengths, indicated by the maxima in the reflection spectra. Four QOMs have been in-
vestigated, with the following wavelengths of electric resonances: 1591 nm (A), 1578 nm (B),
1566 nm (C), and 1546 nm (D). The geometric parameters and the resonance wavelengths
of all QOMs are listed in the Supporting Information (SI). With reducing the size/period
of the pyramids, the resonances shift towards shorter wavelengths. Our experimental obser-
vations are corroborated by numerical simulations with the finite element method (dashed
lines in Fig. 1(d)). The short-wavelength electric resonance, although featuring several mul-
tipole components, is dominated by the electric dipole and quadrupole with the electric
fields mainly in the plane of the metasurface (see the black arrows in Fig. 1(c)). The long-

5
wavelength magnetic resonance features electric field with a more complex structure and is
dominated by the magnetic dipole. More details on the resonances can be found in the SI.
Classical frequency-conversion experiments showed that electric-type resonances with
fields in the metasurface plane can fully utilize the strongest component of the LN second-
(2)
order susceptibility tensor χzzz , leading to improved conversion efficiencies. 32 Furthermore,
because the electric resonances in our metasurfaces have higher field enhancement than the
magnetic resonances (see Fig. 1(c) and the SI), they also provide a larger enhancement of the
density of states. Therefore, for our SPDC experiments pumped at 788 nm we chose QOMs
with electric resonances near the degenerate photon-pair wavelength λdeg = 2 · 788 nm=
1576 nm.
We pumped the QOMs from the substrate side with a continuous-wave (cw) laser at
powers of several tens of mW (Fig. 2(a)). The pump laser was weakly focused using a
parabolic mirror, resulting in a pump beam diameter of 6 µm on the QOM. The photon
pairs generated in the backward direction were collected using the same parabolic mirror.
The filtering system comprised a longpass filter for pump rejection, a bandpass filter of 50 nm
full width at half maximum (FWHM) bandwidth, centered at 1575 nm, and a polarization
analyzer selecting polarization along the LN optics axis. Finally, the generated photons
were coupled into a fiber and registered by two single-photon detectors in a Hanbury Brown-
Twiss (HBT) setup (see Fig. 2(a) and Methods for more details). The collection numerical
aperture (NA), determined by the fiber, was 0.14. Single-photon detection events from the
two detectors were analyzed using a time-to-digital converter.
Figure 2(b) shows coincidence histograms, i.e. the numbers of two-photon detection events
versus the difference in the photon arrival times, measured for ∼ 70 mW pump power over 10
min acquisition time, with the pump polarized along the LN optic axis. The red diamonds
correspond to QOM A and the blue circles to QOM B. The peak in the middle indicates
the simultaneous arrival of photons forming a pair. The coincidences-to-accidentals (peak-
to-background) ratio (CAR) considerably exceeds 2 in both cases, which clearly proves the

6
Figure 2: (a) Correlation experiment. A parabolic mirror focuses a cw pump into the
QOM and collects backward-emitted SPDC. A dichroic mirror separates the SPDC radiation
from the pump, and a 50 nm FWHM bandpass filter centered at 1575 nm transmits nearly
degenerate photon pairs. Another parabolic mirror feeds the SPDC into a Hanbury Brown
– Twiss setup formed by a fiber splitter and two superconducting nanowire single-photon
detectors (SNSPD). (b) Coincidence histograms of degenerate SPDC from QOMs A and B,
shown by red diamonds and blue circles, respectively. The lines are guides to the eye. Gray
stars show the coincidence histogram from an unpatterned LN film of the same thickness as
the nanoresonators. In all measurements the pump power is ∼ 70 mW and the acquisition
time 10 min.

generation of photon pairs in each QOM. 14 The maximal obtained CAR is 361. The rates
of real coincidences, found from the total number of coincidences after subtraction of the
accidental coincidences, are 5.4 ± 0.1 Hz and 1.8 ± 0.1 Hz for QOMs A and B, respectively.
The width of the coincidence histograms is given by the timing jitter of the detectors, which
is about 180 ps.
We compared the photon-pair rates from the QOMs to that of an unpatterned LN film
of the same thickness under the same experimental conditions (coincidence histogram shown
by gray stars in Fig. 2(b)). The peak values of the histograms measured in QOMs A, B
are, respectively, 20 and 7 times higher than for the unpatterned LN. As shown below, we
observe even a stronger enhancement of the pair generation rate by looking into a narrower
spectral range.
To investigate the polarization dependence of SPDC, we have measured the coincidence
rate while rotating the pump polarization angle and keeping the analyzer for signal and idler

7
 3.5

Real coincidences rate [Hz]

3.0
2.5
2.0
1.5
1.0
0.5
0.0
0 50 100 150 200 250 300 350
Pump polarization angle [deg]
Figure 3: Real coincidence rate measured in QOM A versus the pump polarisation angle
with respect to the LN optic axis. The pump power was ∼ 50 mW. The purple curve shows
the theoretical cosine-squared dependence.

photons (Fig. 2(a)) parallel to the LN optic axis. The result for QOM A, at the pump power
∼ 50 mW, is shown in Fig. 3. The pair-generation rate scaled as cos2 θ (purple curve),
θ being the pump polarization angle with respect to the LN optic axis. For the analyzer
oriented orthogonally to the LN optic axis, no photon pairs could be registered regardless
of the pump polarization. This behavior indicates that SPDC was indeed mediated by the
(2)
χzzz tensor component.
Generating entangled photons with tailored spectral correlations is one of the major
tasks in quantum optics. To demonstrate the influence of the resonance on the observed
enhancement of SPDC and its spectral distribution, we measured the spectrum of the
backward-emitted photon pairs via single-photon spectroscopy (SPS). 34 In SPS, the two-
photon wavepacket is spread in time in a dispersive medium, in our case a 1 km-long single-
mode fiber (Corning SMF-28), and then coincidence events between two detectors are regis-
tered in a HBT setup (see Methods for more details). Due to the dispersion, the arrival time
difference between signal and idler photons can be mapped to their spectral separation, and
the spectrum can be retrieved from the coincidence histogram. 13,14 Note that the bandpass
filter (see Fig. 2(a)) was removed in these experiments .
The measured SPDC spectra for QOMs A, B, C, D are plotted in Fig. 4(a). All spectra are

8
 a) b) c)
200 = -15 nm ×128 0 2 xy 0.8
2 xz
150 0.6
100
A 0.4
50 0.2
0 0.0
35 = -2 nm ×22 0 2 xy 1.00
2 xz
30

Normalized SFG intensity [arb. units]

Real coincidences / (hour THz mW)

0.75
25
20 B 0.50
15
10 0.25
5
0 0.00
0.6 = +10 nm ×0.38 0 2 xy
0.15
2 xz
0.5
0.4
0.3 C 0.10
0.2 0.05
0.1
0.0 0.00
= +30 nm ×0.28 0 2 xy
0.08
0.4 2 xz
0.3
0.2
D 0.06
0.04
0.1 0.02
0.0 0.00
1500 1550 1600 1650 1550 1600 1550 1600
Wavelength [nm]
Figure 4: (a) Measured SPDC spectra from QOMs A (red), B (blue), C (orange) and D
(green). Gray points show the SPDC spectrum from the unpatterned LN film. (b) The
spectra obtained through the numerical simulation of SFG at normal incidence. (c) The
SFG spectra calculated for the signal and idler incident at ±2◦ to the normal direction in
the xy (solid lines) and xz (dashed lines) planes.

symmetric around the wavelength of degenerate SPDC, λdeg = 1576 nm, which is because
our experiment cannot distinguish between the signal and idler photons. The measured

9
SPDC spectra are all well localized, in contrast to the spectrum of backward-emitted SPDC
from the unpatterned LN film (gray points in all panels). Here, we observed a much larger
bandwidth, limited only by the spectral sensitivity of our setup. The spectrum of the LN
film is broad because the phase matching is relaxed 14 and the vacuum field seeds SPDC
uniformly over the spectrum. 15
The width of the measured SPDC spectra strongly depends on the detuning ∆λ ≡
λdeg − λER between the degenerate wavelength λdeg and the wavelength λER of the electric
resonance. The latter is marked in each panel by a solid vertical line. Furthermore, the
 −1
2 1
conjugate wavelength of the electric resonance, λconj = λdeg − λER , is marked with a
dashed vertical line. This is the wavelength where the partner photon to the photon emitted
at resonance is detected. We see that each measured spectrum is bounded by the electric
resonance wavelength and its conjugate, whereas no photon pairs are observed at wavelengths
corresponding to the magnetic resonance. This is due to the larger field enhancement of the
electric resonances as well as the direction of their field, which is mainly along the LN optic
(2)
axis and thus uses the χzzz tensor component (Fig. 1(c)). The same effect was observed for
classical frequency conversion in similar LN metasurfaces. 32
The most striking feature of the measured spectra is the giant enhancement of SPDC rate
within a narrow resonance bandwidth. For QOM A, degenerate photon pairs are emitted at a
rate 130 times higher than for the unstructured LN film. The moderate (20-times) advantage
observed in Fig. 2(b) resulted from averaging the spectrum over the 50-nm bandwidth of the
filter, while the resonance is not broader than 10 nm.
One might expect the enhancement to be stronger for QOMs with small detuning of the
electric resonance from degeneracy because efficient pair generation requires a high density of
states at both signal and idler wavelengths. 33 While this is generally true, we see an additional
tendency: the enhancement is stronger for metasurfaces with red-detuned resonances. For
instance, although the resonance of QOM A is more detuned from degeneracy than the one of
QOM B, it is more efficient, whereas QOM C is even less detuned, but very inefficient. This

10
behavior can be explained as follows. According to our linear simulations and measurements
(see the SI), the electric resonance gets blue-shifted under propagation at a nonzero angle
to the metasurface normal. This is important for SPDC in ultrathin sources where photon
pairs are emitted within a broad angle. 35 By collecting SPDC into NA=0.14 we have the
resonances effectively blue-shifted by approximately 10 nm.
To confirm these qualitative considerations, we performed numerical simulations of sum-
frequency generation (SFG) for signal and idler plane waves incident on the sample from the
substrate side. According to the quantum-classical correspondence, 18,26,27 the SFG efficiency
is proportional to the rate of SPDC into the same modes. To obtain a full SFG spectrum,
this calculation was performed for signal and idler wavelengths satisfying energy conservation
(see Methods for more details). In Fig. 4(b) we plot the normalized intensity of SFG emitted
in the backward direction (reversed geometry to the one shown in Fig. 1(a) for SPDC).
The simulations confirm the absence of the magnetic resonance contribution and the de-
pendence of the spectral width on the resonance detuning. Moreover, additional simulations
for SFG from signal and idler incident at moderate angles of ±2◦ (Fig. 4(c)) confirm the
blue shift of the resonance. For QOM A, with detuning ∆λ = −15 nm, the angular tilt
leads to a narrower spectrum with increased count rate, as the resonance is shifting closer to
degeneracy. This is in agreement with the experimental spectrum for QOM A. On the other
hand, for QOM C, with ∆λ = +10 nm, the angular tilt further increases the detuning from
degeneracy, again in accordance with the measurement, where we observe a somewhat wider
spectrum than expected from the normal-incidence resonance wavelengths. This blue shift
also explains why QOM B with ∆λ = −2 nm showed a lower count rate than QOM A with
∆λ = −15 nm. Due to the collection of photons with tilted incidence and the resulting blue
shift of the effective resonance wavelength, the latter gets closer to degeneracy for QOM A
than for QOM B. For QOM D the simulations show good agreement of the spectrum band-
width, but not the shape. We attribute this to the measurement instabilities: this is the
QOM with the weakest SPDC signal, and the measurement was performed over 1 week when

11
the sample might have been slightly displaced and ambient conditions might have changed.
To conclude, we observed photon pairs generated via SPDC in resonant metasurfaces.
Importantly, our experiment was the first-time observation of pronounced (CAR > 2) two-
photon coincidences for SPDC in the reflection geometry. The photon-pair generation was
strongly enhanced by the electric resonance: within its narrow bandwidth, enhancement by
a factor of 130 was measured. Furthermore, the spectral width of the emitted photon pairs
could be controlled through the detuning between the electric resonance and the SPDC de-
generacy wavelength, although the generation efficiency decreases with increasing detuning.
The actual enhancement of SPDC in resonant QOMs is even higher if we take into account
the pump diffraction. Since the QOM period is larger than the pump wavelength, diffraction
into the first orders occurs, the zeroth order carrying only a fraction of the incoming pump
power. 32 Photon pairs can be generated by each diffraction order of the pump; however,
with our NA we only collect pairs corresponding to the zeroth order. We estimate that
the measured photon-pair rate could be considerably increased by using optics with higher
NA or modifying the QOM design. Furthermore, the resonant modes of the QOM could be
optimized to emit generated photon pairs exclusively in the forward or backward direction,
thus enabling more efficient SPDC in either reflection or transmission geometries. 17
Our results are a first step towards the use of nonlinear metasurfaces as versatile sources of
photon pairs. Apart from the spectral control that we demonstrated, QOMs will also enable
far-reaching control of the spatial properties of SPDC, leading to unprecedented possibilities
for the creation of complex two-photon quantum states.

Methods

Correlation experiment in reflection

The pump power was controlled using a half-wave plate (HWP) and a polarizing beam-
splitter (PBS). We used another HWP to rotate the pump polarization. The pump beam

12
with a diameter of ∼ 2.5 mm was focused into the metasurfaces using a 90◦ off-axis gold-
coated parabolic mirror with 15 mm reflective focal length. The same parabolic mirror
collected the backward-generated SPDC radiation. The parabolic mirror played an essential
role as, being free from chromatic aberrations, it ensured the collection of photon pairs from
the same point where the pump laser was focused.
A longpass dichroic mirror with cut-on wavelength at 950 nm split the SPDC radiation
and the portion of the pump laser reflected at the sample surface, with the SPDC being
transmitted. Another longpass filter with cut-on wavelength at 1400 nm filtered out the
remaining pump and removed most part of the fluorescence. We detected degenerate photon
pairs after a 50 nm FWHM bandpass filter centered at 1575 nm. A broadband linear film
polarizer oriented parallel to the LN optic axis selected photon pairs with z polarization.
The polarizer also served as an additional filter removing fluorescence.
A parabolic mirror identical to the one used for focusing and collection fed the radiation
into the input facet of a 50:50 broadband single-mode fiber splitter (1550 nm ± 100 nm).
Due to the identity of the two parabolic mirrors, the NA of the fiber (0.14) also determined
the collection angle of SPDC. The two outputs of the fiber splitter were sent to infrared
superconducting nanowire single-photon detectors (SNSPD). We registered the arrival time
differences between the two detectors using a Swabian Instruments time-tagger (not shown).

Single-photon spectroscopy

To measure the SPDC spectrum, we slightly modified the setup described above. First, we
inserted a 1 km long single-mode fibre (Corning SMF-28) at the input facet of the fiber split-
ter so as to spread the biphoton wavepacket in time. Second, we removed the bandpass filter
so that we could detect a larger bandwidth, from 1400 nm to 1800 nm. We overcame chro-
matic aberrations with the use of parabolic mirrors. Histograms of arrival times differences
were recorded in a similar fashion as in the previous experiment.
Because of the dispersion in the 1 km long fiber, the arrival time differences were redis-

13
tributed over a larger time interval. The delay τ between the signal and idler photons is
given by
τ = L · D(λdeg ) · 2πc(ωs−1 − ωi−1 ), (1)

where L is the length of the fiber, D(λdeg ) is the dispersion parameter [in ps/(nm·km)] of the
fiber at the degenerate wavelength, and the last factor at the right-hand side of Eq. (1) is the
wavelength separation between the signal and idler photons. Due to the energy conservation
condition, Eq. (1) imposes a one-to-one correspondence between τ and the wavelength of
the signal photon. By inverting Eq. (1) we map τ , measured by the time tagger device, to
the wavelength separation, from which the wavelength of the signal photon can be retrieved.
In SMF-28 fiber, D(λdeg = 1576 nm) ∼ 18.8 ps/(nm · km). 36,37 From Eq. (1), the optical
resolution of the system is 8.8 nm, as given by the timing jitter of the detectors and the
length of the fiber.

Simulations of sum-frequency generation

Simulations of sum-frequency generation were done in COMSOL Multiphysics using the un-
depleted pump approximation, and included three steps. The first two are linear simulations
of electromagnetic field for a plane wave excitation at (1) the signal frequency ωs and (2)
the corresponding idler frequency ωi = ωp − ωs , where ωp is the frequency of our pump laser.
Based on the electric fields from the first two steps, we calculated the nonlinear polarization
(see Eq. 2) inside a LN nanoresonator (pyramid and residual layer, see SI) which in turn
served as a source for the final SFG simulation. This algorithm was repeated by varying
the signal frequency and setting the idler frequency accordingly. For the simulations of SFG
with oblique incidence of signal and idler the excitation at the first step (signal study) was
tilted by 2◦ and at the second step (idler study) by −2◦ , both in xy and xz planes (see SI).

14
The components of the nonlinear polarization had the form

PxN L (ωp ) = 2ε0 (d31 [Ex (ωs )Ez (ωi ) + Ez (ωs )Ex (ωi )] − d22 [Ex (ωs )Ey (ωi ) + Ey (ωs )Ex (ωi )]) ,

PyN L (ωp ) = 2ε0 (d22 [Ey (ωs )Ey (ωi ) − Ex (ωs )Ex (ωi )] + d31 [Ey (ωs )Ez (ωi ) + Ez (ωs )Ey (ωi )]) ,

PzN L (ωp ) = 2ε0 (d31 [Ex (ωs )Ex (ωi ) + Ey (ωs )Ey (ωi )] + d33 Ez (ωs )Ez (ωi )) , (2)

where Ex,y,z (ω) were the components of the signal (ωs ) or idler (ωi ) electric field, and d22 =
1.9 pm/V, d31 = −3.2 pm/V, d33 = −19.5 pm/V at 1313 nm. 38
The resulting SFG spectra were multiplied by the SPDC spectrum from the wafer (gray
stars in Fig. 4(a)), to take into account the detectors sensitivity, and convolved with a
Gaussian with FWHM 8.8 nm, to take into account the detectors timing jitter.

Acknowledgement

T. S. C. and M. A. W. are part of the Max Planck School of Photonics supported by

BMBF, Max Planck Society, and Fraunhofer Society. This work was supported by DAAD
GSSP-2017, the German Research Foundation (STA 1426/2-1, PE 1524/13-1, IRTG 2101, SE
2749/1-1, 398816777-SFB1375), the Federal Ministry of Education and Research (13N14147,
03ZZ0434, 13N14877), the Fraunhofer Society within the graduate school of the Leistungszen-
trum Photonik, and the Thuringian State Government within its ProExcellence initiative
(ACP2020).

Supporting Information Available

The Supporting Information is available free of charge via the internet at [Link]
It contains the linear spectra of selected metasurfaces in different geometries (reflection and
transmission) for normal and oblique incidence, the multipole decompositions, and the details
on the field enhancement.

15
References

(1) Yu, N.; Capasso, F. Flat optics with designer metasurfaces. Nature materials 2014, 13,
139–150.

(2) Khorasaninejad, M.; Capasso, F. Metalenses: Versatile multifunctional photonic com-

ponents. Science 2017, 358 .

(3) Zheng, G.; Mühlenbernd, H.; Kenney, M.; Li, G.; Zentgraf, T.; Zhang, S. Metasurface
holograms reaching 80% efficiency. Nature Nanotechnology 2015, 10, 308–312.

(4) Sroor, H.; Huang, Y.-W.; Sephton, B.; Naidoo, D.; Valles, A.; Ginis, V.; Qiu, C.-W.;
Ambrosio, A.; Capasso, F.; Forbes, A. High-purity orbital angular momentum states
from a visible metasurface laser. Nature Photonics 2020, 14, 498–503.

(5) Berzins, J.; Fasold, S.; Pertsch, T.; Bäumer, S. M.; Setzpfandt, F. Submicrometer
nanostructure-based RGB filters for CMOS image sensors. ACS Photonics 2019, 6,
1018–1025.

(6) Zou, C.; Sautter, J.; Setzpfandt, F.; Staude, I. Resonant dielectric metasurfaces: active
tuning and nonlinear effects. Journal of Physics D: Applied Physics 2019, 52, 373002.

(7) Liu, S.; Vabishchevich, P. P.; Vaskin, A.; Reno, J. L.; Keeler, G. A.; Sinclair, M. B.;
Staude, I.; Brener, I. An all-dielectric metasurface as a broadband optical frequency
mixer. Nature Communications 2018, 9, 2507.

(8) Krasnok, A.; Tymchenko, M.; Alù, A. Nonlinear metasurfaces: a paradigm shift in
nonlinear optics. Materials Today 2018, 21, 8 – 21.

(9) Gao, Y.; Fan, Y.; Wang, Y.; Yang, W.; Song, Q.; Xiao, S. Nonlinear holographic all-
dielectric metasurfaces. Nano Letters 2018, 18, 8054–8061.

(10) Solntsev, A. S.; Agarwal, G. S.; Kivshar, Y. S. Metasurfaces for quantum photonics.
[Link] e-Print archive 2020, arXiv:2007.14722 [[Link]] .

16
(11) Poddubny, A. N.; Neshev, D. N.; Sukhorukov, A. A. Quantum nonlinear metasurfaces.
[Link] e-Print archive 2020, arXiv:2008.11035 [[Link]] .

(12) Lee, K. F.; Tian, Y.; Yang, H.; Mustonen, K.; Martinez, A.; Dai, Q.; Kauppinen, E. I.;
Malowicki, J.; Kumar, P.; Sun, Z. Photon-Pair Generation with a 100 nm Thick Carbon
Nanotube Film. Advanced Materials 2017, 29, 1605978.

(13) Okoth, C.; Cavanna, A.; Santiago-Cruz, T.; Chekhova, M. V. Microscale Generation
of Entangled Photons without Momentum Conservation. Phys. Rev. Lett. 2019, 123,
263602.

(14) Santiago-Cruz, T.; Sultanov, V.; Zhang, H.; Krivitsky, L. A.; Chekhova, M. V. Entan-
gled photons from subwavelength nonlinear films. Opt. Lett. 2021, 46, 653–656.

(15) Klyshko, D. N. Photons and Nonlinear Optics; Routledge, 2018.

(16) Poddubny, A. N.; Smirnova, D. A. Nonlinear generation of quantum-entangled pho-

tons from high-Q states in dielectric nanoparticles. [Link] e-Print archive 2018,
arXiv:1808.04811 [[Link]] .

(17) Nikolaeva, A.; Frizyuk, K.; Olekhno, N.; Solntsev, A.; Petrov, M. Directional emission
of down-converted photons from a dielectric nano-resonator. [Link] e-Print archive
2020, arXiv:2011.07842 [[Link]] .

(18) Marino, G. et al. Spontaneous photon-pair generation from a dielectric nanoantenna.

Optica 2019, 6, 1416–1422.

(19) Pertsch, T.; Kivshar, Y. Nonlinear optics with resonant metasurfaces. MRS Bulletin
2020, 45, 210–220.

(20) Liu, S.; Sinclair, M. B.; Saravi, S.; Keeler, G. A.; Yang, Y.; Reno, J.; Peake, G. M.; Set-
zpfandt, F.; Staude, I.; Pertsch, T.; Brener, I. Resonantly Enhanced Second-Harmonic

17
 Generation Using III–V Semiconductor All-Dielectric Metasurfaces. Nano Letters 2016,
16, 5426–5432, PMID: 27501472.

(21) Gili, V. F.; Carletti, L.; Locatelli, A.; Rocco, D.; Finazzi, M.; Ghirardini, L.; Favero, I.;
Gomez, C.; Lemaı̂tre, A.; Celebrano, M.; Angelis, C. D.; Leo, G. Monolithic AlGaAs
second-harmonic nanoantennas. Opt. Express 2016, 24, 15965–15971.

(22) Vabishchevich, P. P.; Liu, S.; Sinclair, M. B.; Keeler, G. A.; Peake, G. M.; Brener, I.
Enhanced Second-Harmonic Generation Using Broken Symmetry III–V Semiconductor
Fano Metasurfaces. ACS Photonics 2018, 5, 1685–1690.

(23) Koshelev, K.; Tang, Y.; Li, K.; Choi, D.-Y.; Li, G.; Kivshar, Y. Nonlinear Metasurfaces
Governed by Bound States in the Continuum. ACS Photonics 2019, 6, 1639–1644.

(24) Anthur, A. P.; Zhang, H.; Paniagua-Dominguez, R.; Kalashnikov, D. A.; Ha, S. T.;
Maß, T. W. W.; Kuznetsov, A. I.; Krivitsky, L. Continuous Wave Second Harmonic
Generation Enabled by Quasi-Bound-States in the Continuum on Gallium Phosphide
Metasurfaces. Nano Letters 2020, 20, 8745–8751, PMID: 33206536.

(25) Liscidini, M.; Sipe, J. E. Stimulated Emission Tomography. Phys. Rev. Lett. 2013, 111,
193602.

(26) Helt, L.; Steel, M. Effect of scattering loss on connections between classical and quantum
processes in second-order nonlinear waveguides. Optics letters 2015, 40, 1460–1463.

(27) Lenzini, F.; Poddubny, A. N.; Titchener, J.; Fisher, P.; Boes, A.; Kasture, S.; Hay-
lock, B.; Villa, M.; Mitchell, A.; Solntsev, A. S.; Sukhorukov, A. A.; Lobino, M. Direct
characterization of a nonlinear photonic circuit’s wave function with laser light. Light:
Science & Applications 2018, 7, 17143–17143.

(28) Timpu, F.; Sendra, J.; Renaut, C.; Lang, L.; Timofeeva, M.; Buscaglia, M. T.;

18
 Buscaglia, V.; Grange, R. Lithium Niobate Nanocubes as Linear and Nonlinear Ul-
traviolet Mie Resonators. ACS Photonics 2019, 6, 545–552.

(29) Gao, B.; Ren, M.; Wu, W.; Hu, H.; Cai, W.; Xu, J. Lithium niobate metasurfaces.
Laser & Photonics Reviews 2019, 13, 1800312.

(30) Ma, J.; Xie, F.; Chen, W.; Chen, J.; Wu, W.; Liu, W.; Chen, Y.; Cai, W.; Ren, M.;
Xu, J. Resonantly tunable second harmonic generation from lithium niobate metasur-
faces. [Link] e-Print archive 2020, arXiv:2002.06594v2 [[Link]] .

(31) Carletti, L.; Zilli, A.; Moia, F.; Toma, A.; Finazzi, M.; De Angelis, C.; Neshev, D. N.;
Celebrano, M. Steering and Encoding the Polarization of the Second Harmonic in the
Visible with a Monolithic LiNbO3 Metasurface. ACS Photonics 2021,

(32) Fedotova, A.; Younesi, M.; Sautter, J.; Vaskin, A.; Löchner, F. J.; Steinert, M.;
Geiss, R.; Pertsch, T.; Staude, I.; Setzpfandt, F. Second-Harmonic Generation in Reso-
nant Nonlinear Metasurfaces Based on Lithium Niobate. Nano Letters 2020, 20, 8608–
8614, PMID: 33180501.

(33) Helt, L. G.; Liscidini, M.; Sipe, J. E. How does it scale? Comparing quantum and
classical nonlinear optical processes in integrated devices. JOSA B 2012, 29, 2199–
2212.

(34) Valencia, A.; Chekhova, M. V.; Trifonov, A.; Shih, Y. Entangled Two-Photon Wave
Packet in a Dispersive Medium. Phys. Rev. Lett. 2002, 88, 183601.

(35) Okoth, C.; Kovlakov, E.; Bönsel, F.; Cavanna, A.; Straupe, S.; Kulik, S. P.;
Chekhova, M. V. Idealized Einstein-Podolsky-Rosen states from non–phase-matched
parametric down-conversion. Phys. Rev. A 2020, 101, 011801.

(36) Zhu, E. Y.; Corbari, C.; Kazansky, P. G.; Qian, L. Self-calibrating Fiber Spectrometer

19
 for the Measurement of Broadband Downconverted Photon Pairs. [Link] e-Print
archive 2015, arXiv:1505.01226 [quant-ph] .

(37) Corning SMF-28 Ultra Optical Fiber. [Link]

coc/documents/Fiber/[Link], (Accessed March 3, 2021).

(38) Shoji, I.; Kondo, T.; Kitamoto, A.; Shirane, M.; Ito, R. Absolute scale of second-order
nonlinear-optical coefficients. JOSA B 1997, 14, 2268–2294.

20

View publication stats