Generation of polarization-entangled Bell states in monolithic photonic

waveguides by leveraging intrinsic crystal properties

Trevor G. Vrckovnik1,2,3 , Dennis Arslan1 , Falk Eilenberger1,2,3 , and Sebastian W. Schmitt1,2,*
1 Fraunhofer Institute for Applied Optics and Precision Engineering IOF, Albert-Einstein-Str. 7, 07745 Jena, Germany
2 Institute of Applied Physics, Abbe Center of Photonics, Friedrich Schiller University Jena, Albert-Einstein-Str. 15, 07745 Jena, Germany
3 Max Planck School of Photonics, Albert-Einstein-Str. 15, 07745 Jena, Germany
* E-mail: sebastian.wolfgang.schmitt@iof.fraunhofer.de

Advanced photonic quantum technologies—from quantum key distribution to quantum computing—require on-chip sources of
entangled photons that are both efficient and readily scalable. In this theoretical study, we demonstrate the generation of
polarization-entangled Bell states in structurally simple waveguides by exploiting the intrinsic properties of nonlinear crystals.
We thereby circumvent elaborate phase-matching strategies that commonly involve the spatial modulation of a waveguide’s
linear or nonlinear optical properties. We derive general criteria for the second-order susceptibility tensor that enable the
generation of cross-polarized photon pairs via spontaneous parametric down-conversion in single-material waveguides. Based
on these criteria, we systematically categorize all birefringent, non-centrosymmetric crystal classes in terms of their suitability.
Using coupled mode theory, we then numerically analyze cuboid waveguides made from two materials that are highly relevant
to integrated photonics: lithium niobate, a well-established platform, and barium titanate, an emerging alternative. We find
that barium titanate consistently outperforms lithium niobate by providing a higher nonlinear efficiency and high concurrence
over a significantly broader spectral range. These findings outline a practical route toward highly efficient, fabrication-friendly,
and scalable sources of polarization-entangled photons for integrated quantum photonic circuits.

1 Introduction The generation of polarization-entangled photon pairs, on

the other hand, remains far more challenging. To date, polar-
The rapid advancements in quantum information process- ization entanglement has only been demonstrated in elaborate
ing and optical telecommunications have significantly in- structures such as Bragg-reflection waveguides21–24 , periodi-
creased the demand for integrated nonlinear quantum pho- cally poled systems with spatial walk-off compensation25,26 ,
tonic devices1,2 . Many emerging quantum technologies—such or hybrid integrated platforms combining multiple optical
as quantum key distribution (QKD) systems3,4 , quantum components and polarization rotators27 . These approaches,
random number generators5,6 , quantum repeaters7,8 , and while effective, introduce fabrication complexity and hinder
photonic quantum processors9,10 —rely on entangled photon scalability.
pairs as a fundamental resource. To enable compact, scalable, We address these technological problems by leveraging the
and stable implementations, it is essential to generate these intrinsic optical properties of nonlinear crystals in conjunction
photons directly on-chip, rather than using bulk crystals or with a minimalist waveguide design: a crystalline cuboid core
optical fibers1,2 . Depending on the application, different types embedded in silicon dioxide (SiO2 ).
of entanglement are required. Polarization entanglement is More precisely, we first establish general requirements on
essential for quantum communication protocols like QKD11 , the χ(2) tensor for the generation of cross-polarized signal
where qubits are encoded in polarization states. Time-bin and idler modes in monolithic waveguides via SPDC, and
entanglement offers robustness against decoherence in fiber then categorize all birefringent, non-centrosymmetric crystal
networks and long-distance transmission12 . Path entangle- classes according to their principal suitability. With this,
ment plays a central role in integrated quantum computing we identify two promising material platforms for integrated
by enabling interference-based operations13 , while frequency photonic devices: lithium niobate28,29 (LiNbO3 ) and barium
entanglement is increasingly used for multiplexed quantum titanate30,31 (BaTiO3 ). Using our own numeric implementa-
communication14 . Each form imposes distinct requirements tion of the coupled mode theory (CMT), capable of handling
on source design and integration, reinforcing the need for arbitrary χ(2) , we perform SHG simulations for various waveg-
versatile and scalable on-chip entanglement generation. uide widths and heights to find the most efficient geometries,
The most widely used physical mechanism for the on-chip where we observe that BaTiO3 achieves higher nonlinear effi-
generation of entangled photons is spontaneous parametric ciencies than LiNbO3 . Finally, SFG simulations of the most
down conversion (SPDC), the quantum analogue of sum fre- promising geometries allow us to infer the likelihood of gener-
quency generation (SFG) and second harmonic generation ated photon pairs and their concurrence in the SPDC process
(SHG)15,16 . For these nonlinear processes to be efficient in via the quantum-classical correspondence principle, where
waveguides, it is essential to utilize a material with a second- BaTiO3 exhibits a high concurrence over a larger bandwidth
order susceptibility tensor χ(2) that has large components at than LiNbO3 .
specific indices, and to achieve phase-matching between the
pump, signal, and idler eigenmodes of the waveguide. The
latter can be accomplished with a careful design of the waveg- 2 Entangled Photon States via SPDC
uide geometry and by utilizing the dispersion and/or bire-
SPDC can be thought of as the time-reversed process of SFG,
fringence of a material’s linear susceptibility χ(1)17 . SPDC
where a single input photon of a higher frequency (known as
has been successfully utilized in integrated waveguides to
the pump), decomposes into two photons of lower frequencies
generate frequency-entangled photon pairs, most notably me-
(known as the signal and idler). Energy conservation of the
diated by quasi-phase matching in domain engineered lithium
process requires that
niobate18,19 and modal phase matching in dispersion engi-
neered semiconductor waveguides20 . ωp = ω s + ω i (1)

1
where ωp , ωs , ωi are the angular frequencies of the pump, 2.4
(a) (b)
signal, and idler photon. As well as this inherent relationship
of the frequencies of the signal-idler pair, there can also be

Refractive index n, Effective refractive index neff

relationships in their polarization, or for waveguides, the BaTiO3 ordinary
2.2
eigenmode they propagate in. For interactions occurring BaTiO3 extraord.
over long distances (such as in waveguides), conservation of SiO2
momentum is also required TM
b c
2 TE
ωp npeff = ωs nseff + ωi nieff (2)
d
where npeff , nseff , and nieff are the effective refractive indices of
(c)
the pump, signal, and idler eigenmodes. If we assume that 1.8
the signal and idler photons can either be predominantly
x-polarized (propagating in the fundamental TE mode) or
predominantly y-polarized (propagating in the fundamental
TM mode), then the two-qubit output state of the SPDC 1.6
process can be expressed as

|ψ⟩ = axx |xx⟩ + axy |xy⟩ + ayx |yx⟩ + ayy |yy⟩ (3)
1.4
where the first term in the ket vector represents the sig- 0.775 1.55
(d)
nal photon polarization, the second term the idler photon Wavelength λ (µm) SiO2
polarization, and the coefficients aij are normalized to: v̂ BaTiO3
(e)
r
⟨ψ|ψ⟩ = |axx | + |axy | + |ayx | + |ayy | = 1
2 2 2 2
(4) ϕ
û
Each of the four basis states in |ψ⟩ represents a different possi-
|E|
ble emission channel for the SPDC process, and entanglement r= y
max(|E|)
occurs when the specific channel used cannot be determined ŵ
0.5 µm
ϕ = ∠ (|Ex | û + |Ey | v̂ + |Ez | ŵ) x
without measurement.
There are four states with maximum entanglement between
Fig. 1 | The role of birefringence. (a) The refractive index of BaTiO3 along the
the signal and idler polarization, known as the Bell states ordinary and extraordinary axis, of SiO2 , and the effective refractive indices of
predominantly x-polarized (TE) or y-polarized (TM) modes. Dashed horizontal
|Φ± ⟩ = √1
2
(|xx⟩ ± |yy⟩) (5) lines illustrate the phase matching condition for the creation of |Ψ± ⟩ Bell states.
(b–d) The electric field profiles of the eigenmodes indicated in (a), where (e) is
|Ψ ⟩ =
±
√1
2
(|xy⟩ ± |yx⟩) (6) the associated color map. The waveguide is made of a BaTiO3 core oriented
such that the extraordinary axis z̃ is parallel to the x-axis of the waveguide. The
core, outlined by dashed lines, has a width of w = 0.61 µm and a height of
where the |Φ± ⟩ Bell states can be created from |ψ⟩ when h = 0.96 µm. The surrounding material is SiO2 .
the coefficients are set to axx = ±ayy and axy = ayx = 0.
Similarly, the |Ψ± ⟩ states can be created when axy = ±ayx
where the angular frequencies represent the weights and are
and axx = ayy = 0.
usually fixed, and the effective refractive indices must be
adjusted to satisfy the equation.
3 Phase Matching Considerations If the signal and idler photons are to be generated in any of
the four basis states of |ψ⟩ (Eq. (3)), we can apply Eq. (7) to
In waveguides, momentum conservation (Eq. (2)) is synony- find the required npeff of the pump mode that achieves phase
mous with phase matching, as the effective refractive indices matching. For the generation of Bell states, two different
determine both the photon momentum and the propagating basis states need to be simultaneously phase-matched. The
phase of the eigenmodes. Since the effective refractive in- required npeff for the |xy⟩ and |yx⟩ basis states are identical
dices usually increase with frequency, it can be expected that for ωs = ωi (rather similar for ωs ≈ ωi ), which means that
phase matching in SPDC must occur between a higher-order the |Ψ± ⟩ Bell states can be phase-matched (rather well) with
pump mode and lower-order signal and idler modes. However, a single pump mode. However, the |xx⟩ and |yy⟩ basis states
since higher-order modes are increasingly more sensitive to will generally require significantly different values of npeff , and
fabrication imperfections and harder to excite in a controlled therefore no single pump mode can be phase-matched to the
manner, the order of all modes should be as low as possible |Φ± ⟩ Bell states. The exception would be if the signal and
to maximize mode overlap and avoid technical difficulties. idler modes are degenerate such that nseff = nieff . However, in
The effective refractive indices can be specifically influenced this case the phase matching condition requires the same npeff
by the waveguide geometry and the linear susceptibility χ(1) for all four basis states, and therefore the |Φ± ⟩ Bell states
of a material, where the orientation of a birefringent material, can only emerge if χ(2) forbids conversion into some of the
as opposed to an isotropic material, represents an additional basis states. Due to these severe fundamental limitations,
degree of freedom in fine-tuning between modes of different we will focus in the remainder of this study on the efficient
polarization. Equations (1) and (2) can also be reinterpreted generation of |Ψ± ⟩ Bell states.
as a weighted arithmetic mean Figure 1a illustrates how the birefringence and aspect ratio
of a BaTiO3 waveguide with a rectangular cross-section can
ωs nseff + ωi nieff
npeff = (7) be exploited to phase-match two fundamental modes of or-
ωs + ωi
thogonal polarization (λ = 1.55 µm; cf. Figs. 1c and 1d) with
a higher-order mode (λ = 0.775 µm; cf. Fig. 1b). This means

2
that the phase matching condition for the |Ψ± ⟩ Bell states If the pump mode is predominantly y-polarized (Exp ≈ 0):
can be fulfilled for a single pump mode (λ = 0.775 µm), but
Px (ωi )
(2) (2) (2)
     
not for the |Φ± ⟩ Bell states. ε0 χxxy χxyy Exs
= · Eyp (18)
Py (ωi )
(2)
2 (2)
χyxy
(2)
χyyy Eys

4 Polarization Entanglement in Waveguides If the signal mode represents the quantum mechanical state
of vacuum fluctuation, meaning the values of Exs and Eys are
The polarization P of an anisotropic and nonlinear material, subject to randomness, but one still wants to enforce that the
whose nonlinear response is approximately independent of idler photon always arises in a polarization state orthogonal
the angular frequency ω of the exciting electric field E, can to the signal photon, which means that the output state is
be represented as17 one of the |Ψ± ⟩ Bell states, then the entries of χ(2) must
∞ have specific values
P(ω) = P(n) (ω) (8)
X
0 ? ↑
 
n=1
 x-polarized pump
↑ ? 0


Pi (ω) = ε0 χij (ω) Ej (ω) (9) χ(2) =  (19)
(1)
X (1)
? ↑ 0

y-polarized pump


? 0
j 
↑
Pi (ω) = ε0 χijk [Ej ∗ Ek ](ω) (10)
(2)
X (2)

jk
where zeros (0) indicates entries that must vanish, question
marks (?) indicate entries that are irrelevant for the case,
where ε0 is the vacuum permittivity, χ(n) the n-th order and arrows (↑) indicate entries that should have large and
susceptibility, and the convolution is defined as: preferably equal absolute values. These symbolic matrices
can be understood as stencils that one must apply to the χ(2)

dω ′
Z
[f ∗ g](ω) = f ω′ g ω − ω′ (11) of a given crystal to check if the crystal and its orientation is


R
2π suitable for the generation of |Ψ± ⟩ Bell states.
Since χ(2) is given in the waveguide reference system, one
We define a waveguide reference system (x, y, z), where the
can not only choose a certain crystal class, but also the orien-
positive z-axis represents the propagation direction of the
tation of the crystal. Crystals of the triclinic, monoclinic, and
waveguide and the positive y-axis points away from the waveg-
orthorhombic crystal systems are biaxial, those of the tetrag-
uide substrate (compare Fig. 2n).
onal, trigonal, and hexagonal crystal systems are uniaxial,
Since low-order modes tend to be mainly transversally
and those of the cubic crystal system are isotropic. In total,
polarized (Ez (ω) ≈ 0), it suffices to focus on the following
there are 18 birefringent and 2 isotropic, non-centrosymmetric
components of Eq. (10)
crystal classes. In the following, we restrict the discussion
[Ex ∗ Ex ](ω) to the birefringent crystal classes, as the isotropic crystal
!
Px (ω)
(2)
 
= ε0 χ (2)
· [Ey ∗ Ey ](ω) (12) classes cannot utilize the phase matching scheme discussed
Py (ω)
(2)
2 [Ex ∗ Ey ](ω) here. Nevertheless, we present a brief adapted analysis of the
isotropic crystal classes in the Supporting Information.
where we shall call In uniaxial crystals, the optical axis corresponds to the
(2) (2) (2) z̃-axis, and birefringence is best utilized when the z̃-axis is
 
χxxx χxyy χxxy
χ(2) := (2) (2) (2) (13) perpendicular to the z-axis. Hence, it is advisable to rotate
χyxx χyyy χyxy
the crystal using the following rotation matrix
the reduced second-order susceptibility in the waveguide
sin(γ̃) cos(γ) sin(γ̃) sin(γ) cos(γ̃)
!
reference system. In the case where the electric field mainly
R(γ, γ̃) = cos(γ̃) cos(γ) cos(γ̃) sin(γ) − sin(γ̃) (20)
consists of three harmonic waves
− sin(γ) cos(γ) 0
E(t) = Re Ep e−iωp t + Es e−iωs t + Ei e −iωi t
(14)


where γ rotates the crystal around the z-axis and γ̃ around
where E , E , E are the eigenmode profiles and ωp , ωs , ωi
p s i its own z̃-axis (see Fig. 2n). The susceptibilities in the waveg-
the angular frequencies of the pump, signal, and idler modes, uide reference system (χ) can thus be represented as linear
the convolution terms evaluated at ω = ωi = ωp − ωs can be combinations of the susceptibilities in the crystal reference
written as system (χ̃):

1 χℓm = (21)
(1)
X (1)
[Em ∗ En ](ωi ) = p
Em Ens + Enp Em
s


δ(ωi − ωp + ωs ) (15) χ̃ij Riℓ Rjm
4 ij

where the overline denotes complex conjugation and δ(·) is χℓmn = (22)
(2)
X (2)
χ̃ijk Riℓ Rjm Rkn
the Dirac delta. Then, by dropping the Dirac delta in favor ijk
of readability, Eq. (12) becomes:
Rotations in π2 -steps are of particular interest here, as they
correspond to a permutation of the coordinate axes, and the
 
Exp Exs
Px (ωi )
(2)
 
ε0 (2) 
= χ · Eyp Eys (16) crystal layers are usually easy to fabricate in these orienta-
Py (ωi ) 2
(2)

Ex Ey + Eyp Exs
p s
tions. Table 1 shows χ(2) for such rotations and reveals that
an additional rotation by ± π2 along γ possesses the following
If the pump mode is predominantly x-polarized (Eyp ≈ 0): symmetry:
Px (ωi ) 0 ? ? 0
(2) (2) (2) π
         
ε0 χxxx χxxy Exs ↑ γ± ↑
= · Exp (17) −→ ∓ 2
(23)
Py (ωi )
(2)
2 (2)
χyxx
(2)
χyxy Eys ↑ ? 0 ? 0 ↑

3
Table 1 | The reduced second-order susceptibility in the waveguide reference Table 2 | The reduced second-order susceptibility in the waveguide reference
system χ(2) for rotations R(γ, γ̃) of birefringent crystals in π 2 -steps, where system χ(2) of BaTiO3 and LiNbO3 for the two crystal orientations used in the
the stacked symbols in front/inside the matrices correspond to those in the simulations. Dots represent vanishing entries. The presented values were taken
row/column headers. Entries of the form ±ijk are shorthand notations for from Weber32 and are valid for SHG pump wavelengths of λ ≈ 1.06 µm.
(2)
±χ̃ijk ; the second-order susceptibility in the crystal reference system. Note that
the intrinsic permutation symmetry ijk = ikj is assumed, while the last row
shows which entries are identical when the Kleinman symmetry applies. χ(2) (pm/V)
γ = γ̃ = − π2 (z̃ ∥ x) γ = γ̃ = 0 (z̃ ∥ y)
(2)
γ̃ = γ̃ = ± π2 13.6 31.4 34.0
  
χ { π0 · · ·
BaTiO3
    · · 34.0 31.4 13.6 ·
±yyy ±yzz yyz ±xxx ±xzz xxz
γ = { π0 ± ± 5.2
  
−68.0 −9.8 · · −9.8
zyy zzz ±zyz
  zxx zzz ±zxz  LiNbO3
 · · −9.8 −9.8 −68.0 ·
zzz zyy ∓zyz zzz zxx ∓zxz
γ = ± π2 ∓ ∓
∓yzz ∓yyy yyz ∓xzz ∓xxx xxz
Kleinm. yyz = zyy xzz = zxz reversed version of the degenerate SPDC process of each of
symm. yzz = zyz xxz = zxx the four basis states in Eq. (3).
The mixed polarization excitation (Figs. 2c, 2f, 2i, 2l) rep-
resents the conversion into the |xy⟩ and |yx⟩ basis states;
In words, this means that if a crystal is suitable for one pump the ones used in the |Ψ± ⟩ Bell states. For both materials,
polarization, it is automatically suitable for the other as well. there are additional branches of high efficiency which are not
The crystal only needs to be rotated by γ ± π2 . Consequently, found in either of the two corresponding single polarization
it suffices to consider only γ = 0 in the following discussion. excitation simulations (Figs. 2a, 2b, 2d, 2e, 2g, 2h, 2j, 2k).
It is also noteworthy that it is beneficial if the Kleinman This indicates that an efficient nonlinear process in these
symmetry applies, since this naturally provides equal values waveguide geometries requires that the signal and idler pho-
for the ↑ entries of the stencils. The χ̃(2) and χ(2) for all tons have orthogonal polarization. In the BaTiO3 simulations
birefringent, non-centrosymmetric crystal classes are listed in (Figs. 2a–2f), these extra branches in the mixed polarization
Table S2 (Supporting Information). Here, we summarize by excitation are more efficient than the equivalent branches
noting that the application of the stencils to χ(2) results in in the LiNbO3 simulations (Figs. 2g–2l). This is because
the following categorization of crystal classes regarding their BaTiO3 has stronger χxyy , χyxy (z̃ ∥ x) or χxxy , χyxx (z̃ ∥ y)
(2) (2) (2) (2)

principal suitability for the generation of |Ψ± ⟩ Bell states: than LiNbO3 , allowing for a more efficient conversion process.
Suitable: 2, m, mm2, 4, 4̄, 4mm, 3m, 6, 6mm In principle, if the system were to be operated in reverse from
Unsuitable: 1, 3, 222, 422, 4̄2m, 32, 6̄, 6̄m2, 622 one of these geometries, i.e. the pump source is the phase-
matched second harmonic mode from these branches, SPDC
Note that in both categories, the actual suitability of a given could generate the |Ψ+ ⟩ Bell state. However, to confirm true
crystal ultimately depends on its specific χ̃(2) values and the Bell state generation, the coefficients aij need to be found,
considered crystal rotations. A non-exhaustive list of suitable so that the output state |ψ⟩ can be properly expressed.
crystals with large χ(2) components is provided in Table S4
(Supporting Information). In the selection of a crystal, one
should further consider technological and socioeconomic fac- 6 SFG Simulations and Concurrence
tors, such as the availability and CMOS-compatibility of the
The concurrence C quantifies the entanglement of arbitrary
involved chemical elements and fabrication processes, as well
two-qubit states33
as their potential health hazards.
C = ⟨ψ|σy ⊗ σy |ψ⟩ (24)
5 SHG Simulations
where |ψ⟩ is a two-qubit state, σy = i 0 one of the Pauli
0 −i

Two promising material platforms for integrated photonic matrices, ⊗ the outer product, and |ψ⟩ the complex conjugate
circuits are lithium niobate28,29 (LiNbO3 , crystal class: 3m) of |ψ⟩. This expression for the concurrence simplifies for the
and barium titanate30,31 (BaTiO3 , crystal class: 4mm), who’s pure state from Eq. (3):
χ(2) are shown in Table 2. By comparison with the stencil
from Eq. (19) it is observed that LiNbO3 has the largest abso- C = 2 |axx ayy − axy ayx | (25)
lute susceptibility component, but in irrelevant positions (?),
The concurrence takes values between 0 and 1, where 1 cor-
while BaTiO3 offers significantly larger components in the
responds to complete entanglement.
relevant positions (↑), making it the preferred material.
The difficulty lies in the determination of the coefficients
To corroborate this analysis, we simulated SHG in waveg-
of the final state |ψ⟩. To avoid the difficulties of a complete
uides made from either BaTiO3 or LiNbO3 with different
quantum-mechanical description, we chose to estimate C via
core dimensions. We also investigated two different crystal
classical wave optics simulations. The key idea is to use the
orientations which exploit the birefringence of the materials
quantum-classical correspondence principle34–36 , where the
to increase the chance of finding phase matched eigenmodes.
probabilities of annihilating a signal-idler photon pair via
The first orientation is given by γ = γ̃ = − π2 (see Eq. (20)
SFG is the same as generating said pair via SPDC.
and Fig. 2n), which aligns the crystal’s z̃-axis with the waveg-
SFG simulations allow us to determine the concurrence as
uide’s x-axis. In the following, we will refer to this orientation
a function of the signal (and idler) frequency. If we perform
as z̃ ∥ x. The second orientation is given by γ = γ̃ = 0, which
four separate SFG simulations, in each of which we excite a
we will analogously refer to as z̃ ∥ y.
pair of modes |ij⟩ (where i, j ∈ {x, y}) with a power P0 =
In Fig. 2, we show that by changing the excitation polar-
P (0, ωs ) + P (0, ωi ) and measure the power Pij = P (ℓ, ωp ) of
ization of the SHG simulations, we can simulate the time-
the generated pump mode at the waveguide end z = ℓ (where
4
 BaTiO3 LiNbO3
z̃ ∥ x z̃ ∥ y z̃ ∥ x z̃ ∥ y (m) Color map

SHG efficiency η̃ (% W-1 cm-2 )

103
(a) (d) 1.6 (g) (j)
1.2 102
Height h (µm)

1.4
1 101
x

1.2
0.8 100
1
0.6 10-1
Excitation polarization

(b) (e) 1.6 (h) (k) -1 0 1

1.2 Polarization fraction PF
Height h (µm)

1.4
1 (n)
y

1.2
0.8
SiO
2
1
0.6
1.6 Cry
(c) (f) (i) (l) sta
1.2 l
Height h (µm)

h
1.4
x+y

1 w γ y
1.2
z̃
0.8
1 ỹ
0.6
0.6 0.8 1 1.2 0.6 0.8 1 1.2 1 1.2 1.4 1.6 1 1.2 1.4 1.6
x
z γ̃ x̃
Width w (µm) Width w (µm) Width w (µm) Width w (µm)

Fig. 2 | (a–l) SHG efficiency and polarization fraction, encoded via the (m) two-dimensional color map, for (a–f) BaTiO3 and (g–l) LiNbO3 waveguides. (n) The crystal
reference system (x̃, ỹ, z̃) is rotated with respect to the waveguide reference system (x, y, z), where γ rotates the crystal around the z-axis and γ̃ around its own
z̃-axis. The crystalline waveguide with width w and height h is fully embedded in SiO2 . In (a–c, g–i), the crystal is rotated by γ = γ̃ = − π 2 , which means z̃ ∥ x.
In (d–f, j–l), the crystal is rotated by γ = γ̃ = 0, which means z̃ ∥ y. The waveguides are excited by (a, d, g, j) the x-polarized, (b, e, h, k) the y-polarized, or (c, f, i, l) both
x- and y-polarized fundamental eigenmodes. The widths and heights marked by (c, f, i, l) circles and (c) a square relate to Fig. 3 and Fig. 1, respectively.

Pij keeps record of the excitation), we can expect in turn for for that basis state to decrease. This creates a certain band-
the SPDC process that pij = Pij /P0 is the probability for width around the degenerate frequency which supports highly
finding the signal-idler pair |ij⟩ associated with the coefficient entangled photon pair generation. The exact bandwidth is
aij at the end of the waveguide. Hence, the coefficients of dependent on the geometry, material, and dimensions of the
the final state |ψ⟩ can be estimated as waveguide, however it is important to note that these simula-
√ tions show that each of these geometries are capable of being
pij
|aij | ≈ √ (26) used as integrated Bell state sources near the degenerate
pxx + pxy + pyx + pyy frequency they were designed for.
where we note that this follows the same normalization re-
quirement from Eq. (4). 7 Conclusion
Once the four coefficients are known, the concurrence can
easily be calculated using Eq. (25). We see that for the |Ψ± ⟩ This study demonstrates that it is possible to efficiently gen-
Bell states, which have axy = ±ayx = √12 and axx = ayy = 0, erate polarization-entangled Bell states via SPDC in struc-
the concurrence is equal to 1, showing that these are states turally simple waveguides by leveraging intrinsic crystal prop-
with maximum entanglement. erties. Compared to other phase-matching techniques, our
In Fig. 3, we show the results of the SFG sweeps of different approach may require fewer and less complicated fabrica-
geometries previously identified to show promise for |Ψ± ⟩ tion steps, thereby offering a higher integration density and
Bell state generation in both BaTiO3 and LiNbO3 . For robustness to fabrication tolerances.
each waveguide, the pump photon wavelength was fixed at To this end, we first identified the components of χ(2)
λ = 775 nm, while the signal and idler photon wavelengths that enable the generation of cross-polarized photon pairs in
were varied. After a waveguide length of ℓ = 100 µm, the four monolithic waveguides. We then assumed crystal rotations
coefficients for the output state were found using Eq. (26), and that maximize the effect of birefringence in uniaxial crystals
the corresponding concurrence was calculated with Eq. (25). and are generally easy to fabricate. With this, we categorized
For each geometry, the concurrence at the degenerate fre- all relevant crystal classes regarding their principal suitability
quency is 1 and this then begins to decrease as the signal and found barium titanate to be superior to lithium niobate,
photon frequency is offset from the degenerate frequency. as it offers a higher nonlinear efficiency and a high concurrence
This makes sense as at the degenerate frequency the signal over a broader bandwidth.
and idler photons are indistinguishable, and therefore |axy | In summary, our findings highlight the potential for highly
and |ayx | must always be equal. As the signal photon fre- efficient and scalable on-chip sources of polarization-entangled
quency increases or decreases, the effective refractive index photons, paving the way for technological advancements in
of the signal and idler photons may change at different rates. quantum key distribution and quantum computing technolo-
This causes the phase matching to get worse between the gies. This work not only provides a practical framework for
pump, signal, and idler modes, resulting in the coefficient
5
 BaTiO3 LiNbO3
z̃ ∥ x z̃ ∥ y z̃ ∥ x z̃ ∥ y

(a) (1.05, 0.83) (e) (1.10, 0.78) (i) (0.98, 0.99) (m) (0.94, 1.17)
1

(b) (1.35, 1.36) (f) (1.30, 1.07) (j) (1.13, 1.60) (n) (1.31, 0.98)
1

(c) (0.83, 1.14) (g) (0.79, 1.02) (k) (1.14, 1.14) (o) (1.62, 1.56)
1

(d) (1.12, 1.34) (h) (1.20, 1.27) (l) (1.61, 1.25) (p) (1.66, 1.53)
1

0
0.8 1 1.2 1.4 1.6 0.8 1 1.2 1.4 1.6 0.8 1 1.2 1.4 1.6 0.8 1 1.2 1.4 1.6
Signal frequency ωs (PHz) Signal frequency ωs (PHz) Signal frequency ωs (PHz) Signal frequency ωs (PHz)

|axx | |axy | |ayx | |ayy | C

Fig. 3 | (a–p) Normalized coefficients |axx | , |axy | , |ayx | , |ayy | of the four basis states in the two-qubit state |ψ⟩ as a function of the signal photon frequency ωs ,
calculated from the efficiency of the conversion into the pump mode after the corresponding SFG simulation, and the concurrence C of the output state. The waveguide
cores were made of (a–h) BaTiO3 or (i–p) LiNbO3 , with the crystal rotated such that (a–d, i–l) z̃ ∥ x or (e–h, m–p) z̃ ∥ y. Each waveguide was simulated over a length
of ℓ = 100 µm. The tuple in each panel shows the waveguide width and height (w, h) in micrometers, and corresponds to a circular mark in Fig. 2.

future developments in integrated photonic circuits for quan- plane, βm is the propagation constant of the mode, and z is
tum applications, but also contributes to the understanding the current position in the propagation direction.
of nonlinear optical processes in waveguides. Using Eq. (27) as an ansatz in the vector wave equation,
we find a set of coupled differential equations that describe
Methods how each mode amplitude evolves as the mode propagates
along the waveguide37 :
Simulation Methods
∂ ω
Am (z) = i ϵ0 e−iβm z
From Maxwell’s curl equations the vector wave equation can ∂z Z4
be derived, the solutions of which are a series of eigenvectors
with corresponding eigenvalues. Each eigenvector describes Em (x, y, ω) · PNL (x, y, z, ω) dx dy (29)
R2
how an electric and magnetic field can propagate in the
waveguide, and the corresponding eigenvalue is known as We then find the eigenmodes of a specific waveguide geome-
the propagation constant. The combined electric and mag- try and numerically solve the corresponding set of coupled
netic field is known as an eigenmode, and once normalized, differential equations from Eq. (29) as an initial value prob-
eigenmodes of a specific frequency have the unique property lem. This results in a list of complex mode amplitudes, each
that they are all orthonormal to each other. Therefore, any evaluated at each z-step of the initial value solver. With
electromagnetic field propagating through a waveguide can these amplitudes, the total E and H fields at each frequency
be decomposed into a linear superposition of eigenmodes as can be found in the waveguide using Eqs. (27) and (28).
For the electric field from Eq. (14), the total guided power
E(x, y, z, ω) = Am (z) Em (x, y, ω) eiβm z (27)
X
P (z) at a point z in the waveguide can be decomposed into
m the contributions P (z, ωq ) of the modes q ∈ {p, s, i} as
H(x, y, z, ω) = Am (z) Hm (x, y, ω) eiβm z (28)
X
P (z) = P (z, ωp ) + P (z, ωi ) + P (z, ωs ) (30)
m
1
Z
where Am is the amplitude of the eigenmode, Em and Hm P (z, ωq ) = Re Eq (r, ωq ) × Hq (r, ωq ) · ẑ dx dy (31)


are the electric and magnetic mode profiles in the transverse 2 R2

6
Since SPDC is a purely quantum process, it cannot be sim- [4] H. Yu et al., “Quantum key distribution implemented
ulated directly. However, we can instead simulate the time- with d-level time-bin entangled photons”, Nat.
reversed processes, namely SHG and SFG, and from those Commun. 2025, 16, 1.
results infer the results of the SPDC process. In these time- [5] M. Herrero-Collantes, J. C. Garcia-Escartin, “Quantum
reversed processes, the signal and idler photon become the random number generators”, Rev. Mod. Phys. 2017, 89,
input to the simulation, while the pump photon becomes the 015004.
output. Without loss of generality, it can be assumed that the
waveguide starts at z = 0 and is excited there with a certain [6] Y. Liu et al., “Device-independent quantum
signal power P (0, ωs ) and idler power P (0, ωi ). The normal- random-number generation”, Nature 2018, 562, 548.
ized efficiency η with which pump photons are generated is [7] Z. Zhao et al., “Experimental Realization of
then defined as Entanglement Concentration and a Quantum Repeater”,
P (z, ωp ) Phys. Rev. Lett. 2003, 90, 207901.
η(z) := (32)
P (0, ωi ) P (0, ωs ) [8] K. Azuma, K. Tamaki, H.-K. Lo, “All-photonic
quantum repeaters”, Nat. Commun. 2015, 6, 1.
with units of [η] = W−1 . Since P (z, ωp ) is generally propor-
[9] N. K. Langford et al., “Efficient quantum computing
tional to P (0, ωi ) P (0, ωs ) z 2 , i.e. dependent on the square of
using coherent photon conversion”, Nature 2011, 478,
the waveguide length, we can further define a normalized
360.
efficiency η̃
[10] K. Alexander et al., “A manufacturable platform for
5 photonic quantum computing”, Nature 2025, 641, 876.
Z ℓ Z ℓ
η̃ := arg min 2 2
dz = 5 z η(z) dz
2
(33)
 
η(z) − az
a 0
ℓ 0 [11] A. Poppe et al., “Practical quantum key distribution
with polarization entangled photons”, Opt. Express
with units of [η] = W−1 m−2 , where η̃z 2 fits the normalized
2004, 12, 3865.
efficiency η(z) in the least squares sense and ℓ is the actual
length of the waveguide. Values of equivalent quantities are [12] E. Fitzke et al., “Scalable Network for Simultaneous
often given in literature in units of % W−1 cm−2 . Pairwise Quantum Key Distribution via
Moreover, we define the polarization fraction PF as Entanglement-Based Time-Bin Coding”, PRX Quantum
2022, 3, 020341.
|Ex (x, y, ω)|2 − |Ey (x, y, ω)|2 dx dy
R
PF(ω) = R2
(34) [13] M. Krenn et al., “Entanglement by Path Identity”,
|E(x, y, ω)|2 dx dy Phys. Rev. Lett. 2017, 118, 080401.
R
R2

which indicates whether the guided electric field is predom- [14] S. Wengerowsky et al., “An entanglement-based
inantly x-polarized (PF = 1) or predominantly y-polarized wavelength-multiplexed quantum communication
(PF = −1). network”, Nature 2018, 564, 225.
[15] W. H. Louisell, A. Yariv, A. E. Siegman, “Quantum
Acknowledgments Fluctuations and Noise in Parametric Processes. I.”
Phys. Rev. 1961, 124, 1646.
The authors acknowledge the Fraunhofer Attract Grant SILI- [16] M. H. Rubin et al., “Theory of two-photon
QUA No. 40-04866, the BMBF projects MEXSIQUO Grant entanglement in type-II optical parametric
No. 13N16967 and SINNER Grant No. 16KIS1792, and the down-conversion”, Phys. Rev. A 1994, 50, 5122.
Collaborative Research Center (CRC/SFB) 1375 NOA.
[17] R. W. Boyd, “Nonlinear optics”, Elsevier, 2020.
[18] J.-L. Tambasco et al., “Domain engineering algorithm
Conflict of Interest for practical and effective photon sources”, Opt. Express
The authors declare no conflict of interest. 2016, 24, 19616.
[19] C. J. Xin et al., “Spectrally separable photon-pair
Data Availability Statement generation in dispersion engineered thin-film lithium
niobate”, Opt. Lett. 2022, 47, 2830.
The data that support the findings of this study are available [20] M. Placke et al., “Telecom-Band Spontaneous
from the corresponding author upon reasonable request. Parametric Down-Conversion in AlGaAs-on-Insulator
Waveguides”, Laser Photonics Rev. 2024, 18, 9.
References [21] S. V. Zhukovsky et al., “Generation of
maximally-polarization-entangled photons on a chip”,
[1] E. Pelucchi et al., “The potential and global outlook of Phys. Rev. A 2012, 85, 013838.
integrated photonics for quantum technologies”, Nat.
Rev. Phys. 2021, 4, 194. [22] A. Orieux et al., “Direct Bell States Generation on a
III-V Semiconductor Chip at Room Temperature”,
[2] G. Moody et al., “2022 Roadmap on integrated Phys. Rev. Lett. 2013, 110, 160502.
quantum photonics”, J. Phys.: Photonics 2022, 4,
012501. [23] R. T. Horn et al., “Inherent polarization entanglement
generated from a monolithic semiconductor chip”, Sci.
[3] T. Jennewein et al., “Quantum Cryptography with Rep. 2013, 3, 1.
Entangled Photons”, Phys. Rev. Lett. 2000, 84, 4729.

7
[24] S. Francesconi et al., “On-chip generation of hybrid
polarization-frequency entangled biphoton states”,
Photonics Res. 2023, 11, 270.
[25] L. Sansoni et al., “A two-channel, spectrally degenerate
polarization entangled source on chip”, npj Quantum
Inf. 2017, 3, 1.
[26] X. Zhang et al., “Polarization-entangled photon pairs
generation from a single lithium niobate waveguide with
single poling period”, Opt. Express 2025, 33, 24193.
[27] N. Matsuda et al., “A monolithically integrated
polarization entangled photon pair source on a silicon
chip”, Sci. Rep. 2012, 2, 1.
[28] A. Boes et al., “Lithium niobate photonics: Unlocking
the electromagnetic spectrum”, Science 2023, 379,
6627.
[29] Y. Hu et al., “Integrated electro-optics on thin-film
lithium niobate”, Nat. Rev. Phys. 2025, 7, 237.
[30] A. A. Demkov et al., “Materials for emergent
silicon-integrated optical computing”, J. Appl. Phys.
2021, 130, 7.
[31] A. Raju et al., “High-Q Monolithic Ring Resonators in
Low-Loss Barium Titanate on Silicon”, Laser Photonics
Rev. 2025.
[32] M. J. Weber, “Handbook of Optical Materials”, CRC
Press, 2018.
[33] W. K. Wootters, “Entanglement of Formation of an
Arbitrary State of Two Qubits”, Phys. Rev. Lett. 1998,
80, 2245.
[34] L. G. Helt, M. Liscidini, J. E. Sipe, “How does it scale?
Comparing quantum and classical nonlinear optical
processes in integrated devices”, J. Opt. Soc. Am. B
2012, 29, 2199.
[35] A. Eckstein et al., “High-resolution spectral
characterization of two photon states via classical
measurements: High-resolution spectral characterization
of two photon states via classical measurements”, Laser
Photonics Rev. 2014, 8, L76.
[36] G. Kulkarni et al., “Classical model of spontaneous
parametric down-conversion”, Phys. Rev. Res. 2022, 4,
033098.
[37] T. Suhara, M. Fujimura, “Waveguide Nonlinear-Optic
Devices”, Springer Berlin Heidelberg, 2003.

8
Supporting Information:
Generation of polarization-entangled Bell states in monolithic photonic
waveguides by leveraging intrinsic crystal properties
Trevor G. Vrckovnik1,2,3 , Dennis Arslan1 , Falk Eilenberger1,2,3 , and Sebastian W. Schmitt1,2,*
1 Fraunhofer Institute for Applied Optics and Precision Engineering IOF, Albert-Einstein-Str. 7, 07745 Jena, Germany
2 Institute of Applied Physics, Abbe Center of Photonics, Friedrich Schiller University Jena, Albert-Einstein-Str. 15, 07745 Jena, Germany
3 Max Planck School of Photonics, Albert-Einstein-Str. 15, 07745 Jena, Germany
* E-mail: sebastian.wolfgang.schmitt@iof.fraunhofer.de

S1 Contracted Kleinman notation

To enhance the readability in the following tables, we employ
the contracted Kleinman notation1 χijk = χiℓ , where the
(2) (2)

indices x, y, z are enumerated by 1, 2, 3 and the last pair of

indices jk maps to ℓ according to:
γ y
jk 11 22 33 23, 32 13, 31 12, 21
(a) y (b)
(S1) z̃
ℓ 1 2 3 4 5 6
ỹ
S2 Second-order susceptibility of birefringent
ỹ x
and non-centrosymmetric crystals x z x̃
z γ̃ x̃
Table S1 indicates how the entries of χ(2) permute when ϕ
the crystal is rotated by R(γ, γ̃) in steps of π2 (cf. Fig. S1a). z̃
Table S2 shows the general structure of both χ̃(2) and χ(2)
Fig. S1 | The coordinate systems and rotations used for (a) birefringent and
for all birefringent and non-centrosymmetric crystal classes. (b) isotropic crystal classes. The crystal reference system (x̃, ỹ, z̃) lies rotated
Table S4 lists the values of χ(2) for selected crystals and in the waveguide reference system (x, y, z). In (a), γ rotates the crystal around
the z-axis and γ̃ around its own z̃-axis. In (b), ϕ rotates the crystal around the
rotations R(γ = 0, γ̃). Note that the values for other rotations z-axis and, in the illustrated case, the permutation matrix P in R(ϕ) was set to
can be deduced from Table S1. the identity matrix I.

S3 Second-order susceptibility of isotropic and

non-centrosymmetric crystals
The cubic crystal class, and some of the birefringent crystal
classes (e.g. 222 and 4̄2m), can still be used for the genera-
tion of |Ψ± ⟩ Bell states if the crystal is rotated in another
manner. For example, the following rotation matrix R(ϕ) is
particularly well suited for the cubic crystal class

cos(ϕ) sin(ϕ)
 
√1 √1 1
√
2 2 2
R(ϕ) = P ·  − √12 cos(ϕ) − √12 sin(ϕ) 1
√
2
 (S2) Table S1 | The reduced second-order susceptibility in the waveguide reference
sin(ϕ) − cos(ϕ) 0 system χ(2) for rotations R(γ, γ̃) of birefringent crystals in π
2 -steps, where
the stacked symbols in front/inside the matrices correspond to those in the
where P is an arbitrary permutation matrix that sets one (2)
row/column headers. Entries of the form ±iℓ are shorthand notations for ±χ̃iℓ ;
of the (110) , (011) , (101) crystal planes perpendicular to the the second-order susceptibility in the crystal reference system. The last row
shows which entries are identical when the Kleinman symmetry applies.
z-axis, and ϕ rotates the selected crystal plane round the
z-axis (cf. Fig. S1b).
Table S3 shows that the two applicable cubic crystal classes χ(2) (γ, γ̃) γ̃ = { π0 γ̃ = ± π2
(23, 4̄3m) are, when rotated by R(ϕ), well suited for the 
±22 ±23 24 ±11 ±13 15
  
generation of |Ψ± ⟩ Bell states. Lastly, Table S5 shows selected γ = { π0 ± ±
32 33 ±34 31 33 ±35
isotropic crystals with high values for χ̃14 .
(2)
33 32 ∓34 33 31 ∓35
   
γ = ± π2 ∓ ∓
∓23 ∓22 24 ∓13 ∓11 15
References
Kleinman 24 = 32 13 = 35
[1] R. W. Boyd, “Nonlinear optics”, Elsevier, 2020. symmetry 23 = 34 15 = 31
[2] M. J. Weber, “Handbook of Optical Materials”, CRC
Press, 2018.

9
Table S2 | The second-order susceptibility in the crystal reference system χ̃(2) , and in reduced form in the waveguide reference system χ(2) , for all birefringent and
(2)
non-centrosymmetric crystal classes. Entries of the form ±iℓ are shorthand notations for ±χ̃iℓ . Dots represent vanishing entries. The reduced form χ(2) is shown
for selected crystal rotations R(γ = 0, γ̃), where the stacked symbols inside the matrices correspond to those in the column headers. The reduced form for other
values of γ can be derived from Table S1 via symmetry considerations. The last row shows which entries are identical when the Kleinman symmetry applies.

Crystal Crystal χ̃(2) χ(2) (γ = 0, γ̃)

system class γ̃ = { π0 γ̃ = ± π2
(2)
±iℓ → ±χ̃iℓ
11 12 13 14 15 16
! 
Triclinic 24 15
  
±22 ±23 ±11 ±13
1 C1 21 22 23 24 25 26
(biaxial) 32 33 ±34 31 33 ±35
31 32 33 34 35 36
· · · 14 · 16
! 
Monoclinic
  
±22 ±23 · · · ·
2 C2 21 22 23 · 25 ·
(biaxial) · · ±34 · · ·
· · · 34 · 36
11 12 13 · 15 ·
! 
24 15
  
· · ±11 ±13
m C1h · · · 24 · 26
32 33 · 31 33 ±35
31 32 33 · 35 ·
Ortho- · · · 14 · ·
!    
· · · · · ·
rhombic 222 D2 · · · · 25 ·
· · · · · ·
(biaxial) · · · · · 36
· · · · 15 ·
! 
24 15
  
· · · ·
mm2 C2v · · · 24 · ·
32 33 · 31 33 ·
31 32 33 · · ·
· · · 14 15 ·
! 
Tetragonal 15 15
  
· · · ·
4 C4 · · · 15 −14 ·
(uniaxial) 31 33 · 31 33 ·
31 31 33 · · ·
· · · 14 15 ·
! 
15
  
· · −15 · ·
4̄ S4 · · · −15 14 ·
−31 · · 31 · ·
31 −31 · · · 36
· · · 14 · ·
!    
· · · · · ·
422 D4 · · · · −14 ·
· · · · · ·
· · · · · ·
· · · · 15 ·
! 
15 15
  
· · · ·
4mm C4v · · · 15 · ·
31 33 · 31 33 ·
31 31 33 · · ·
· · · 14 · ·
!    
· · · · · ·
4̄2m D2d · · · · 14 ·
· · · · · ·
· · · · · 36
11 −11 · 14 15 −22
! 
Trigonal 15 15
  
±22 · ±11 ·
3 C3 −22 22 · 15 −14 −11
(uniaxial) 31 33 · 31 33 ·
31 31 33 · · ·
11 −11 · 14 · ·
!    
· · · ±11 · ·
32 D3 · · · · −14 −11
· · · · · ·
· · · · · ·
· · · · 15 −22
! 
15 15
  
±22 · · ·
3m C3v −22 22 · 15 · ·
31 33 · 31 33 ·
31 31 33 · · ·
11 −11 · · · −22
! 
Hexagonal
  
±22 · · ±11 · ·
6̄ C3h −22 22 · · · −11
(uniaxial) · · · · · ·
· · · · · ·
· · · · · −22
!    
±22 · · · · ·
6̄m2 D3h −22 22 · · · ·
· · · · · ·
· · · · · ·
6 C6 see: 4
622 D6 see: 422
6mm C6v see: 4mm
Kleinman symmetry 12 = 26, 13 = 35, 14 = 25 = 36, 15 = 31, 16 = 21, 23 = 34, 24 = 32

10
Table S3 | The second-order susceptibility in the crystal reference system χ̃(2) , and in reduced form in the waveguide reference system χ(2) , for all isotropic and
(2)
non-centrosymmetric crystal classes. Entries of the form ±iℓ are shorthand notations for ±χ̃iℓ . Dots represent vanishing entries. The reduced form χ(2) is shown for
selected rotations R(ϕ), where the stacked symbols inside the matrices correspond to those in the column headers. Note that the presented rotations are independent
of the permutation matrix P.

Crystal Crystal χ̃(2) χ(2) (ϕ)

system class ϕ = { π0 ϕ = ± π2
(2)
±iℓ → ±χ̃iℓ
· · · 14 · ·
!    
· · ±14 · ∓14 ·
Cubic 23 T · · · · 14 ·
±14 · · · · ∓14
· · · · · 14
4̄3m Td see: 23

11
Table S4 | The reduced second-order susceptibility of selected birefringent crystals
in the waveguide reference system χ(2) , scaled by a factor of 12 , for selected
crystal rotations R(γ = 0, γ̃). The presented values were taken from Weber2
and are valid for SHG pump wavelengths of λ ≈ 1.06 µm.

Crystal Chemical χ (γ
1 (2)
2
= 0, γ̃)
class formula (pm/V) γ̃
 
· · −5.4
mm2 PbNb2 O11 0
−5.9 −8.9 ·
5.9
 
· · π
6.5 −8.9 · 2

8.6
 
· ·
LiInO2 0
8.6 15.8 ·
9.9
 
· · π
9.9 15.8 · 2

7.6
 
· ·
KTiOPO4 0
5.0 13.7 ·
6.1
 
· · π
6.5 13.7 · 2

11.3
 
· ·
KNbO3 0
11.3 −19.6 · Table S5 | The second-order susceptibility of selected isotropic crystals in the
(2)
crystal reference system χ̃14 , scaled by a factor of 12 . The presented values were
 
· · −12.9 π
2
−12.9 −19.6 · 2 taken from Weber and are valid for the indicated SHG pump wavelengths λ.
 
· · −12.8
Ba2 NaNb5 O15 0 Crystal Chemical 1
χ̃ Wavelength
(2)
−12.8 −17.6 · 2 14

· · −12.8
 class formula (pm/V) λ (µm)
π
−12.8 −17.6 · 2 4̄3m CdTe 167 10.60
GaAs 209 1.06
 
· · −26.3
4̄ InPS4 0 GaP 72 1.06
−26.3 · ·
GaSb 628 10.60
26.3

· ·
InAs 364 1.06
π

 26.3 · · 2
InP 143 1.06
6.0

· ·
4mm SrBaNb5 O15 0 InSb 520 1.06
 4.3 11.3 ·
ZnTe 90 1.06
33.3

· ·
PbTiO3 0
 37.6 7.5 ·
17.0

· ·
BaTiO3 0
15.7 6.8 ·
1.7
 
· −1.1
3m LiTaO3 0
−1.1 −16.4 ·
 
· · −1.1 π
−1.1 −16.4 · 2

2.6
 
· −4.9
LiNbO3 0
−4.9 −34.0 ·
 
· · −4.9 π
−4.9 −34.0 · 2
 
· · −5.0
6 LiIO3 0
−5.0 −5.2 ·
28.9
 
· ·
6mm CdS 0
−26.4 44.0 ·
8.0
 
· ·
SiC 0
8.6 −14.4 ·
6.7
 
· ·
α-ZnS 0
−7.6 13.8 ·

12