General simulation method for spontaneous parametric down- and parametric

up-conversion experiments
Felix Riexinger,1, 2, ∗ Mirco Kutas,1, 2 Björn Haase,1, 2 Patricia
Bickert,1 Daniel Molter,1 Michael Bortz,1 and Georg von Freymann1, 2
1
Fraunhofer Institute for Industrial Mathematics ITWM,
Fraunhofer-Platz 1, 67663 Kaiserslautern, Germany
2
Department of Physics and Research Center OPTIMAS,
Technische Universität Kaiserslautern (TUK), 67663 Kaiserslautern, Germany
Spontaneous parametric down-conversion (SPDC) sources are an important technology for quan-
tum sensing and imaging. We demonstrate a general simulation method, based on modeling from
first principles, reproducing the spectrally and spatially resolved absolute counts of a SPDC ex-
periment. By simulating additional processes and effects we accomplish good agreement with the
arXiv:2112.07243v1 [quant-ph] 14 Dec 2021

experimental results. This method is broadly applicable and allows for the separation of contributing
processes, virtual characterization of SPDC sources, and enables the simulation of many quantum
based applications.

Entangled and correlated photon pairs have become ahertz range and signal photons in the visible range.
the basis for many applications in quantum optics. They The extreme wavelength spread between signal and idler
are used in various quantum based schemes such as ghost leads to a setup that covers a large range in frequency
imaging [1, 2], optical coherence tomography [3], spec- and emission directions and further has multiple quasi-
troscopy [4, 5], quantum sensing [6], and imaging with phasematching (QPM) orders overlapping in the same
undetected photons [7–9]. One of the most prominent wavelength range. In the terahertz range additional
methods for generating entangled photon pairs is spon- processes such as parametric up- and (nonspontaneous)
taneous parametric down-conversion (SPDC), where a down-conversion occur parallel to SPDC. In order to ad-
pump laser photon decays into two lower-energy photons equately match the experimental data we include these
in a medium with a second-order nonlinearity. The en- processes in our simulation.
tanglement can exist in many of the photon properties, The high qualitative and quantitative agreement with
such as polarization, wavelength, or momentum, making experimental results demonstrates the capabilities of our
SPDC a versatile source for entangled photons. Addi- simulation method even for complex SPDC sources.
tionally, the process is easy to implement and well under- Our model is based on the second-order nonlinear in-
stood experimentally. However, an accurate simulation teraction of electromagnetic fields together with a first-
method for SPDC sources is required as the foundation order perturbation theory approximation. We start with
for simulations of many applications with entangled pho- the Hamiltonian [13]
tons. The availability of detailed simulations can help Z
1 (2)
resolve problems such as finding limitations to the res- HNL (t) = ε0 drζjkl (r)Dj (r, t)Dk (r, t)Dl (r, t), (1)
olution or visibility of quantum imaging [8, 9]. This is- 3
sue is becoming more relevant as applications for SPDC (2)
where ζjkl is the second-order inverse susceptibility
sources are on the verge of a breakthrough, but their lim- tensor and D are the displacement fields. The formu-
iting factors need to be understood better to exploit the lation with the displacement fields instead of the electric
full potential of such quantum based applications. fields is necessary to ensure consistency after quantiza-
The theory of SPDC is well developed and multiple tion [14, 15].
approaches to simulating the properties of the created We describe the pump beam as a classical monochro-
photons are available [10, 11]. However, these SPDC sim- matic Gaussian beam with linear polarization. We as-
ulations are often limited to the paraxial regime [11], or sume that the pump is undepleted, meaning its power
do not predict absolute photon conversion rates [12]. In is large enough such that it is not significantly reduced
this letter we propose and demonstrate a novel simulation due to nonlinear conversion events. In addition, we use
method for SPDC sources and the subsequent measure- the approximation of a collimated beam such that the
ment setup. The sparse use of approximations makes the curvature and the Gouy phase can be neglected. The
underlying model applicable to a wide range of SPDC pump propagates along the z-axis, which we define par-
sources from the ultraviolet to the terahertz regime. allel to the optical axis of the system. The signal and
Our method reproduces the spectrally and spatially idler fields are described using a plane-wave decomposi-
resolved absolute photon count rates. This is demon- tion separated into positive and negative frequency com-
strated on an experiment with idler photons in the ter- ponents D̂+ and D̂− with
r
+
X ε0 n2k ~ωk
D̂ (r, t) = i âk,σ k,σ ei(k·r+ωk t) , (2)
∗ [email protected] 2V
k,σ
 2

and D̂− being the hermitian conjugate of D̂+ . Here, refractive indices is considered following [18]. The spec-
V is the quantization volume, âk,σ is the annihilation tral dependencies are especially relevant since we simu-
operator for a photon with momentum k, and k,σ is the late a large spectral range in the terahertz regime, from
direction of the displacement field vector indexed with ∼ 0.1 THz to 3.6 THz. The spatial variation cannot be
the polarization σ. neglected here, because the transverse momentum con-
We then calculate the two-photon state |ψ(t)i using servation dictates a large emission angle range for the
first-order perturbation theory: idler photons.
With the idler in the terahertz range, at room temper-
i t
 Z 
ature, thermal photons at the idler wavelength have to
|ψ(t)i ≈ 1 − dτ HNL (τ ) |ψ(0)i, (3)
~ 0 be taken into account [19].
These thermal photons interact with the pump laser as
where the initial state |ψ(0)i is a vacuum state. Under well. Through parametric down-conversion, additional
the assumption that the quantization volume is large, we photons at the signal wavelength are created. We derive
can make a transition from sums to integrals in Eq. (2). this process analogously to SPDC, where instead of an
From this we obtain the signal count rate density: initial vacuum state we have a thermal state for the idler.
This leads to the same phasematching properties, but
1 instead of a ”1” from the vacuum fluctuations we obtain
Γd (ks ) = hψ(TI )|↠(ks )â(ks )|ψ(TI )i (4) the thermal fluctuations
TI
X X Z 2 1
=Z dki3 kA(ks , ki )k , (5) Nth = , (8)
exp(~ωi /kB Tc ) − 1
σs ,σi m odd

with where Tc is the temperature of the crystal. The influence

of thermal fluctuations for the signal can be neglected.
The emitted spectrum is propagated from the crystal
(2) r   onto a detector through an optical system. The num-
χ ωs ωi 1
A(ks , ki ) = eff sinc ∆k z L ber of counts for a single detector pixel is obtained by
m n2s n2i 2
    integrating over all signal wave vectors that are propa-
1 1 gated to this pixel. However, not all photons created in
× exp − (∆kx2 + ∆ky2 )wp2 sinc ∆ωTI (6)
4 2 the crystal arrive at the detector. We consider the losses
from internal reflection in the crystal, during propaga-
and tion through the optical setup and the efficiency of the
detector summarized into a single factor η. With this
we obtain the photon counts for the detector pixel with
16P wp2 L2 TI
Z= . (7) indices (i, j)
(2π)7 ε0 np c
Z
(i,j)
Here TI is the interaction time for a pump photon with Rd = ηT (1 + Nth ) dks Γd (ks ), (9)
Ω(i,j)
the nonlinear medium, m denotes the odd QPM orders,
and we sum over permutations of the indices j, k, l and
where T is the illumination time of the detector and
substitute new indices to denote pump (p), signal (s) and
Ω(i, j) is defined such that every ray with kΩ ∈ Ω(i, j) at
idler (i). ∆kz = kpz − ksz − kiz + kΛ and ∆kj = ksj + kij
the crystal exit is propagated through the optical system
with j = x, y are the longitudinal and transversal phase
onto the detector pixel (i, j). Along the same lines we
mismatches, and ∆ω = ωp − ωs − ωi corresponds to the obtain the rate for the parametric up-conversion process,
energy mismatch. The width of the transverse part is de- where a thermal photon and a laser photon merge into a
termined by the waist radius wp of the pump beam. The signal photon. The only changes are the signs of the ter-
periodic poling offset kΛ = 2mπ/Λ depends on the poling
ahertz frequency ωi in the terms ∆ω and the wave vector
period of the crystal Λ and the QPM order m. P denotes
kij in the respective ∆kj terms with j = x, y, z. Up-
the power of the pump beam and L the length of the
(2) conversion is only caused by thermal fluctuations, such
crystal. The refractive indices n and χeff are functions that the up-conversion rate is given by:
of the wave vectors kp , ks , ki and their corresponding
polarizations. The spectral dependence of the nonlinear
Z
coefficient is modeled with Miller’s rule [16]. The spatial Ru(i,j) = ηT Nth dks Γu (ks ). (10)
Ω(i,j)
(2)
variation of χeff is considered by calculating the effective
value from the tensor components and displacement field A sketch of the model for the experimental setup is
directions [17]. The spectral dependence of the refractive shown in Fig. 1, where only the simulated components
indices is modeled with Sellmeier equations in the visi- are shown. The setup consists of a narrow bandwidth
ble range and polynomial fits to experimental data from laser, a nonlinear crystal and an imaging system. The
literature in the terahertz range. Spatial variation in the spatial and spectral components of the created signal
 3

vertical slit grating

d = 1.1 mm 1908 lines/mm the previous method, but assumed a constant value. Sec-
pump laser ond, we added the full optical model to the simulation,
659.58 nm PPLN
f1 = 150 mm f2 = 125 mm f3 = 400 mm
which changes the shape of the spectrum. The value of
(2)
χ311 = 49 pm/V at 661 nm and 0.75 THz is a scaled value
(a) (b) from [21]. The contributions of this parameter are small
such that a reliable fit is not possible. The refractive in-
sCMOS dices, for ∼ 5 mol.% MgO-doped LiNbO3 , are taken from
camera M [24, 25].
The propagation through the optical system is mod-
eled by paraxial ray optics assuming ideal optical com-
ponents. The paraxial approximation is reasonable as
only the signal photons are detected. The idler photons
FIG. 1. Layout of the experimental setup. Only the compo- which are emitted at larger angles are not considered.
nents used in the simulations are included. f: lens. PPLN: Since the longitudinal positions of the optical compo-
periodically poled MgO-doped LiNbO3 -crystal. M: mirror. nents are not exactly known, they are estimated from
Length of optical paths are not to scale. (a) represents a the imaging properties of the setup. This is done us-
cross section of the spectrum after the slit. (b) sketches the ing an optimization algorithm to minimize the squared
separation of spectral components by the grating. The ratio difference of three experimentally measured values: the
da /db quantifies the asymmetry introduced in this step. transformation of wavelength into a position on the x-
axis, the relation between emission angle and position on
the y-axis, and the squeezing of a monochromatic circular
beam, defined by the ratio da /db . We also penalize de-
photons are separated by the imaging system and im-
viations from the measured positions. This procedure of
aged onto the detector. The imaging system contains a
estimating the parameters of the setup mimics the align-
slit to limit the transmission of rays with large kx , shown
ment process performed in the experiment, where the
in Fig. 1 (a), resulting in a sharper image on the detector.
components are moved around their nominal positions
This effect is shown in Fig. 1 (b), where the remaining
to obtain a sharper image. The calculated values deviate
parts of the ellipses are separated along the θs axis. A
from the nominal distances of the setup, but reproduce
transmission grating separates the spectral components
the imaging properties of the actual experiment. Further
of the photons. Photons with a fixed wavelength are
sources of errors in the optical system such as misalign-
emitted in a cone shape, shown in Fig. 1 (a). This leads
ment transverse to the optical axis or derivations from
to an imperfect separation of spatial and spectral com-
nominal values are not included in our model.
ponents, which is shown by the full and partial ellipses in
Fig. 1 (b). The ratio da /db is a measure of how much an We employ a Monte-Carlo integration scheme to eval-
ellipse is squeezed. As it approaches zero, the ellipses are uate the integrals in Eq. (4), approximating each squared
imaged as vertical lines, making the spread in θs large sinc functions with the sum of three scaled Gaussian func-
and the one in λs small. The system further contains tions. This allows for efficient sampling while maintain-
several filters and Bragg gratings to suppress the pump ing a small error in the approximated function. Other
radiation that are not considered in this model. A de- options such as a single Gaussian or a Cosine Gauss ap-
tailed description of the experimental realization is given proximation [26] are either less accurate or less efficient.
by Haase et al. [20]. The laser is modeled with a wave- The sum in Eq. (4) is evaluated up to the seventh QPM
length of λp = 659.58 nm and a beamwaist wp = 43 µm. order. Since positive and negative orders are possible,
The nonlinear crystal is a periodically poled MgO-doped a total of eight summands are evaluated. Higher orders
lithium niobate crystal with dimensions 5 × 1 × 10 mm3 contribute less than 70 counts per pixel to the spectrum
(H ×W ×L) and a poling period of 170 µm. Due to sym- in the observed range and are therefore neglected. We
metries in the nonlinear susceptibility only the values of consider both ordinary and extraordinary polarization
(2) (2) for signal and idler. Type 0 parametric conversion con-
the χ333 and χ311 components need to be considered. The
(2) tributes most of the spectrum in the investigated range,
χ222 component is not relevant to any processes in our type I contributes less than 165 counts per pixel in the
(2)
setup. We use a value of χ333 = 327 pm/V at 661 nm and region above 663.7 nm and less than 2 counts per pixel
0.75 THz which we obtained from a fit to the experimen- elsewhere. Type II contributes less than 1 count per pixel
(2)
tal results. The fitted χ333 value agrees with a scaled in the whole range.
value from the infrared range [21] but is ∼ 1.5 times The experimental and numerical results for the full
larger than other scaled values from the terahertz range spectrum are shown in Fig. 2. The experimental spec-
[22, 23]. The difference can be explained by different fac- trum is a single image as recorded by the sCMOS camera
tors for the Hamiltonian ( 21 and 13 ) used in the classical with the average dark count rate subtracted. Simulated
and quantum approaches. Compared to our previous es- and experimental spectrum show four distinct tails for
timation [20] this value is two times larger. There are up- as well as down-conversion. The contributing QPM
two reasons for this: First, we did not use Miller’s rule in orders for each tail are given in panel (b) of the figure.
 4

8000

Emission angle θs (deg)

2
(a)
1 7000

0
6000

−1

Counts per pixel

5000
−2

656 658 660 662 664

4000
Emission angle θs (deg)

2
(b)
3000
1

0
2000

−1

1000
−2
1 -3
m=5 3 -5 -3 -1 -1 1 -5 3 5
656 658 660 662 664
Wavelength λs (nm)

FIG. 2. Experimental (a) and simulated (b) frequency-angular spectrum. The average dark count rate is subtracted from the
experimental spectrum. The dashed lines in (a) denote the location of the cuts presented in Fig. 3 and 4. The numbers m in
(b) indicate the contributions of the different QPM orders. The tails on the left correspond to up-conversion, the ones on the
right to down-conversion.

The tails corresponding to the fifth QPM order are barely 7000
visible, while the contributions of the seventh order can- s = 0.2° Experiment + s
6000 Experiment - s
Counts per pixel

not be distinguished at all due to the lower conversion 5000 Simulation ± s

efficiency of higher orders. The experiment is limited
4000
to a scattering angle of around ±2.3◦ by the apertures
of the optical components. The simulated spectrum ex- 3000
tends beyond this angle as the limiting apertures are not 2000
included in the model. 1000
The simulation shows more counts than the experi- 0
ment at wavelengths lower than 655 nm and higher than 655.58 657.58 659.58 661.58 663.58
Wavelength s (nm)
663.5 nm for up- and down-conversion, respectively. Po-
tential reasons are limiting apertures in the experiment
or an incorrect model for the terahertz refractive indices. FIG. 3. Horizontal cut through the experimental and simu-
The idler photons in this range have a frequency of over lated spectra at θs = 0.2◦ . The positive and negative sign
2.5 THz which is beyond the measured range of our ref- for θs correspond to cuts through the upper and lower half of
erence, covering frequencies from 0.3 THz to 1.9 THz. Fig. 2 respectively. The simulated spectrum is symmetric in
(2) θs , thus only one line is shown.
The value of χeff depends on the refractive indices in
our model such that an overestimation of n increases the
counts.
Figures 3 and 4 show horizontal and vertical cuts of of the peaks. The experimental spectrum shows some
the spectrum. The positive and negative values for θs remaining pump light around the laser wavelength at the
correspond to the sign given in Fig. 2. As the simulated center of Fig. 3. This is more pronounced for the +θs
spectrum is symmetric in θs only one line is shown. Both case. In the model we assume the pump to be blocked
cuts show good agreement in position, width and height completely, therefore the simulated spectrum does not
 5

4000 tween the two experimental results. Peaks found in the

Counts per pixel (a) s = 656.8 nm cuts from larger angles or larger wavelengths match in
3000 shape and position. The peaks of the simulated spec-
trum are significantly higher for this region. Note that
2000 the simulated spectrum exhibits some minor modulation
along the θs -axis which are numerical artifacts caused by
1000 the employed sampling method.
The demonstrated simulation of parametric conversion
0 spectra shows good agreement with the experiment in
0.0 0.5 1.0 1.5 2.0 2.5 spectral and angular distributions as well as absolute
5000 photon counts. Qualitative and quantitative features
(b) s = 662.0 nm of the experimentally obtained spectrum can be repro-
4000
Counts per pixel

duced. Various effects such as SPDC, the influence of

thermal photons and parametric up-conversion were sim-
3000 ulated. This shows the potential of applying our model to
2000 predict accurate characteristics of photon sources. And
the possibility of identifying the contributions of differ-
1000 ent processes. The model can be simplified for faster
0 computation times. An important step to improve the
simulation results is to use better estimates for the crys-
0.0 0.5 1.0 1.5 2.0 2.5
Emission Angle s (deg)
tal characteristics in the investigated frequency range.
The method provides significant benefits over traditional
methods such as the calculation of phasematching curves
FIG. 4. Vertical cuts through the experimental and simulated as no information about the width or intensity of the
spectra. Panel (a) shows a cut in the up-conversion regime, spectrum is provided there. Our numerical method also
panel (b) in the down-conversion regime. The color coding is allows for reconstruction of the spectrum at the crystal
the same as in Fig. 3. face which allows to investigate spatial, spectral and cor-
relation properties without the need of building optical
setups for measuring them. Compared to previous results
show the pump rest. [20], the propagation of the spectrum through the mea-
The tails of the experimental spectrum are slightly surement setup improves the simulation results. Further
narrower and more pronounced, which is due to the ex- research is needed to separate influences of model errors
periment being adjusted for a maximally sharp image, in the up- and down-conversion model from those in the
while the simulated setup is not. It is optimized to re- optical measurement setup. Nonetheless, the presented
produce three imaging characteristics of the experiment. model provides the necessary basis for the simulation of
Observed differences between simulation and experiment many quantum optic applications such as quantum imag-
are of the same order of magnitude as the differences be- ing.

[1] T. B. Pittman, Y. H. Shih, D. V. Strekalov, and A. V. and G. von Freymann, Terahertz quantum sensing, Sci-
Sergienko, Optical imaging by means of two-photon ence Advances 6, eaaz8065 (2020).
quantum entanglement, Phys. Rev. A 52, R3429 (1995). [7] G. B. Lemos, V. Borish, G. D. Cole, S. Ramelow, R. Lap-
[2] M. J. Padgett and R. W. Boyd, An introduction to ghost kiewicz, and A. Zeilinger, Quantum imaging with unde-
imaging: quantum and classical, Philosophical Transac- tected photons, Nature 512, 409 (2014).
tions of the Royal Society A: Mathematical, Physical and [8] J. Fuenzalida, A. Hochrainer, G. B. Lemos, E. Ortega,
Engineering Sciences 375, 20160233 (2017). R. Lapkiewicz, M. Lahiri, and A. Zeilinger, Resolution
[3] A. Vanselow, P. Kaufmann, I. Zorin, B. Heise, H. M. of quantum imaging with undetected photons, arXiv
Chrzanowski, and S. Ramelow, Frequency-domain opti- preprint arXiv:2010.07712 (2020).
cal coherence tomography with undetected mid-infrared [9] M. Gilaberte Basset, A. Hochrainer, S. Töpfer,
photons, Optica 7, 1729 (2020). F. Riexinger, P. Bickert, J. R. León-Torres, F. Steinlech-
[4] C. Lindner, J. Kunz, S. J. Herr, S. Wolf, J. Kießling, ner, and M. Gräfe, Video-rate imaging with undetected
and F. Kühnemann, Nonlinear interferometer for fourier- photons, Laser & Photonics Rev. 15, 2000327 (2021).
transform mid-infrared gas spectroscopy using near- [10] J. Schneeloch and J. C. Howell, Introduction to the
infrared detection, Opt. Express 29, 4035 (2021). transverse spatial correlations in spontaneous paramet-
[5] M. Kutas, B. Haase, J. Klier, D. Molter, and G. von Frey- ric down-conversion through the biphoton birth zone, J.
mann, Quantum-inspired terahertz spectroscopy with Opt. 18, 053501 (2016).
visible photons, Optica 8, 438 (2021). [11] S. Trajtenberg-Mills, A. Karnieli, N. Voloch-Bloch,
[6] M. Kutas, B. Haase, P. Bickert, F. Riexinger, D. Molter, E. Megidish, H. S. Eisenberg, and A. Arie, Simulating
 6

correlations of structured spontaneously down-converted Millim. Terahertz Waves 32, 1144 (2011).
photon pairs, Laser & Photonics Rev. 14, 1900321 [20] B. Haase, M. Kutas, F. Riexinger, P. Bickert, A. Keil,
(2020). D. Molter, M. Bortz, and G. von Freymann, Spontaneous
[12] R. Ramı́rez-Alarcón, H. Cruz-Ramı́rez, and A. B. U’Ren, parametric down-conversion of photons at 660 nm to the
Effects of crystal length on the angular spectrum of spon- terahertz and sub-terahertz frequency range, Opt. Ex-
taneous parametric downconversion photon pairs, Laser press 27, 7458 (2019).
Phys. 23, 055204 (2013). [21] D. N. Nikogosyan, Nonlinear optical crystals: a complete
[13] J. Schneeloch, S. H. Knarr, D. F. Bogorin, M. L. survey (Springer Science & Business Media, 2006) pp.
Levangie, C. C. Tison, R. Frank, G. A. Howland, M. L. 41–44.
Fanto, and P. M. Alsing, Introduction to the absolute [22] J. Hebling, A. G. Stepanov, G. Almási, B. Bartal, and
brightness and number statistics in spontaneous para- J. Kuhl, Tunable THz pulse generation by optical recti-
metric down-conversion, J. Opt. 21, 043501 (2019). fication of ultrashort laser pulses with tilted pulse fronts,
[14] M. Hillery and L. D. Mlodinow, Quantization of electro- Appl. Phys. B 78, 593 (2004).
dynamics in nonlinear dielectric media, Phys. Rev. A 30, [23] K. Vodopyanov, Optical THz-wave generation with
1860 (1984). periodically-inverted GaAs, Laser & Photonics Rev. 2,
[15] N. Quesada and J. E. Sipe, Why you should not use the 11 (2008).
electric field to quantize in nonlinear optics, Opt. Lett. [24] X. Wu, C. Zhou, W. R. Huang, F. Ahr, and F. X.
42, 3443 (2017). Kärtner, Temperature dependent refractive index and
[16] R. C. Miller, Optical second harmonic generation in absorption coefficient of congruent lithium niobate crys-
piezoelectric crystals, Appl. Phys. Lett. 5, 17 (1964). tals in the terahertz range, Opt. Express 23, 29729
[17] R. W. Boyd, Nonlinear Optics, 3rd ed. (Academic Press, (2015).
Inc., USA, 2008) pp. 40–46. [25] O. Gayer, Z. Sacks, E. Galun, and A. Arie, Temperature
[18] M. Born and E. Wolf, Principles of Optics: Electromag- and wavelength dependent refractive index equations for
netic Theory of Propagation, Interference and Diffraction MgO-doped congruent and stoichiometric LiNbO3 , Appl.
of Light, 60th ed. (Cambridge University Press, 2019) pp. Phys. B 91, 343 (2008).
806–807. [26] P. Kolenderski, W. Wasilewski, and K. Banaszek, Mod-
[19] G. K. Kitaeva, S. Kovalev, A. Penin, A. Tuchak, and eling and optimization of photon pair sources based on
P. Yakunin, A method of calibration of terahertz wave spontaneous parametric down-conversion, Phys. Rev. A
brightness under nonlinear-optical detection, J. Infrared 80, 013811 (2009).