PPLN Λ = 18.99 µm · 1550 → 775 nm · 2 px/µm

2.7 Waveguides in TFLN

L1 Phenomenon

rib waveguide of Li et al. 2026, x-cut, 1550 nm

Thin-film LN on insulator (LNOI, TFLN) is a crystalline film, 300 to 900 nm thick in one supplier's range V, bonded on a silica layer. Partial etching leaves a rib on a thinner slab: Li et al. 2026 use a 300 nm film, 120 nm slab and 3.5 µm rib reported. The mode fills a cross-section of order λ2, the index contrast to the silica cladding allows tight bends, and electrodes sit micrometres from the guide (6 µm gap in Li et al. 2026 reported).

Propagation loss: 0.27 dB/cm on 4- and 6-inch wafers patterned by deep-UV lithography reported; 1.3 dB/m in dry-etched rings (intrinsic Q 29 million) reported; Q above 1e8 in chemo-mechanically polished rings reported. The material limit of annealed ion-sliced film is estimated at 0.2 dB/m (Q 1.6e8) reported.

Rib of Li et al. 2026, x-cut, to scaleengine
6 µm gapw 3.5 µmt 300 nme 180 nmslab 120 nmSiO₂ 2 µmsidewall 60°AuAuSiO₂ claddingburied SiO₂, 4.7 µmSiairz (c)1 µm
film t / etch e / slab300 / 180 / 120 nmreported
top width w3.5 µmreported
sidewall, reference not stated60°reported
cladding, PECVD SiO22 µmreported
electrode gap, Au6 µmreported
neff, TE, 1550 nm1.7696engine
ng, TE, 1550 nm2.135engine
Drawn from the engine model of the reported geometry: sidewall 60° from the substrate plane, Au 1.0 µm on the slab, buried SiO2 4.7 µm (broken). One scale for both axes.
Rib modes: neff vs top width, TE0 / TM0 / TE1full-vector FEM engine
0.60.811.5231.761.781.801.821.841.861.88top width w (µm)neffoutsidelatticeslab TEslab TM
TE0TM0TE1flaggedslab, film t derived◇ Li et al. 2026 rib enginehatch: outside lattice
400 nm
0.60
1.75 µm
Sidewall
λ
Cut
Readout mode
Preset
neff–.––––ng–.–––
Aeff–.–– µm²LN fraction–.––
TE fraction–.––status
Li et al. 2026 rib, w 3.5 µm (outside the width axis), TE0, direct solve: neff 1.7696, ng 2.135 engine
Held-out error, median / p95clean / flaggedneff cleanneff flaggedneff allng cleanng flaggedng all
TE093 / 329.9e−4 / 4.1e−34.5e−4 / 2.3e−38.4e−4 / 4.1e−33.8e−4 / 5.7e−36.1e−4 / 1.2e−24.5e−4 / 7.4e−3
TM079 / 359.3e−4 / 6.4e−31.3e−3 / 7.0e−31.1e−3 / 6.7e−34.0e−4 / 1.4e−24.0e−3 / 1.6e−28.6e−4 / 1.4e−2
TE147 / 661.1e−3 / 4.3e−31.4e−3 / 4.1e−31.1e−3 / 4.2e−39.0e−4 / 1.4e−22.4e−3 / 2.4e−21.5e−3 / 1.6e−2
Full-vector FEM (Nedelec/Lagrange order 2), eo-modulator-atlas engine, commit eb97f96, GPL-3.0-or-later: 6300 solves on the grid t 300 to 900 nm, etch fraction 0.2 to 1.0, w 0.6 to 3.0 µm (geometric), sidewall 60/70/80°, λ 0.775/1.064/1.55 µm, x- and z-cut; multilinear interpolation, λ on the grid only. Stack: LN film on SiO₂ with SiO₂ cladding over the rib; no Si substrate (substrate leakage not modelled), no electrodes. LN: Zelmon 1997 congruent, 21 °C; SiO₂: Malitson 1965. Held-out set: 126 random geometries solved directly. engine
c axis of the film6 µm3.5 µmAuAuSiO₂, 2 µmburied SiO₂SiLN 300 nm1 µm
Film / etch / slab
300 / 180 / 120 nm
reported
Top width
3.5 µm
reported
Sidewall
60°
reported
PECVD SiO2 cladding
2 µm
reported
Au electrode gap
6 µm
reported
neff, TE, 1550 nm
1.7696
engine
ng, TE, 1550 nm
2.135
engine

To scale. x-cut film, light along y, c across the gap: TE light and the electrode field both lie along c. Sidewall reference direction not stated in the paper; drawn 60° from the substrate plane as in the engine model, Au 1.0 µm thick on the slab, buried oxide (4.7 µm) broken. engine

c (z), in planew 1.75 µm1 µmSiO₂SiO₂ cladding0.60.811.5231.761.781.801.821.841.861.88top width w, µmneffoutsidelatticeslab TEslab TM
400 nm
0.60
1.75 µm
Sidewall
Wavelength
Cut
Mode in the readout
neff
–.––––
ng
–.–––
Aeff
–.–– µm²
LN fraction
–.––
TE fraction
–.––
Status
Li et al. 2026 rib, w 3.5 µm, TE0
1.7696 / 2.135
engine

Ellipse: |Ex| and |Ey| shares of the readout mode (TE fraction), not a polarization state; light at 1064 and 1550 nm in false color, dashed. Grey band on a curve: flagged lattice points; dotted lines: slab of thickness t derived; diamond: Li et al. 2026 engine value, outside the width axis.

Held-out error, median / p95clean / flaggedneff cleanneff flaggedneff allng cleanng flaggedng all
TE093 / 329.9e−4 / 4.1e−34.5e−4 / 2.3e−38.4e−4 / 4.1e−33.8e−4 / 5.7e−36.1e−4 / 1.2e−24.5e−4 / 7.4e−3
TM079 / 359.3e−4 / 6.4e−31.3e−3 / 7.0e−31.1e−3 / 6.7e−34.0e−4 / 1.4e−24.0e−3 / 1.6e−28.6e−4 / 1.4e−2
TE147 / 661.1e−3 / 4.3e−31.4e−3 / 4.1e−31.1e−3 / 4.2e−39.0e−4 / 1.4e−22.4e−3 / 2.4e−21.5e−3 / 1.6e−2

Full-vector FEM (Nedelec/Lagrange order 2), eo-modulator-atlas engine, commit eb97f96, GPL-3.0-or-later: 6300 solves on the grid t 300 to 900 nm, etch fraction 0.2 to 1.0, w 0.6 to 3.0 µm (geometric), sidewall 60/70/80°, λ 0.775/1.064/1.55 µm, x- and z-cut; multilinear interpolation, λ on the grid only. Stack: LN film on SiO₂ with SiO₂ cladding over the rib; no Si substrate (substrate leakage not modelled), no electrodes. LN: Zelmon 1997 congruent, 21 °C; SiO₂: Malitson 1965. Held-out set: 126 random geometries solved directly. engine

L2 Quantitative model

x-cut vs z-cut, ring anisotropy, loss vs Q

Geometry. Film thickness t, etch depth e (slab t − e), top width w, sidewall angle; cladding air or oxide. Li et al. 2026: t = 300 nm, e = 180 nm, slab 120 nm, w = 3.5 µm, sidewall 60° (reference direction not stated), 2 µm PECVD SiO2 cladding reported; engine neff = 1.7696, ng = 2.135 at 1550 nm engine.

x-cut vs z-cut

x-cut, propagation along yz-cut
TE indexne 2.1376 at 1550 nm derivedno 2.2111 derived, same in every in-plane direction
TM indexnone
EO, in-plane electrodesTE uses r33TE uses r22; TM with vertical field uses r33
χ(2)TE type 0 uses d33TM uses d33
ringsTE index varies around the ringTE isotropic in plane
polingcoplanar in-plane electrodes, field nonuniform with depthvertical field through the stack

Indices: Zelmon Sellmeier, congruent LN, 21 °C. Coefficient mapping: 2.2 (r tensor) and 2.3 (d tensor).

Anisotropy around a ring (x-cut, bulk approximation). For TE propagating at angle φ from crystal z within the film plane, the in-plane polarization sees

1n(φ)2=cos2φno2+sin2φne2

varying between ne (propagation along y) and no (propagation along z): 2.1376 to 2.2111 at 1550 nm derived, mode confinement ignored. The local phase velocity, the QPM condition and the r33 projection all vary twice per round trip.

Ring: in-plane TE index n(φ), bulk approximationderived
2.142.162.182.202.22z (c)yφ
Wafer
 30°
1.55 µm
n(φ)2.1920Δn in plane0.0736
no2.2111ne2.1376
n(φ)none
Square: current φ. Radial axis offset (inner circle at the lowest labelled index). Zelmon CLN, 21 °C, 0.40 to 5.00 µm; film indices may differ.
c (z)yxprobene 2.1376no 2.2111
 30°
1.55 µm
Wafer cut
n(φ) at the probe
2.1920
no
2.2111
ne
2.1376
In-plane span
0.0736

Band brightness = TE index. Two bright and two dark arcs per round trip turn with the crystal; the probe stays at the top, where φ is the angle between the light and c (x-cut) or x (z-cut). Bulk approximation, mode confinement ignored; ring not to scale. Indices: Zelmon Sellmeier, congruent LN, 21 °C, 0.40 to 5.00 µm; film indices may differ. derived

Mode hybridization. Where a TE and a TM mode of the same waveguide would cross in neff, any broken symmetry (slanted sidewalls, a slab on one side, an off-axis crystal orientation) couples them into hybrid modes with an anticrossing (hybridization). In TFLN ridges this affects even the fundamental modes (Pan et al. 2019). In x-cut strip waveguides the hybrid regions depend on dimensions, sidewall angle and wavelength, and tapers through them convert TM0 to higher-order TE (99.5%, numerical) reported. In x-cut microrings the rotating birefringence converts TE to TM and creates avoided mode crossings that suppress one four-wave-mixing sideband reported. Arrayed-waveguide gratings need anisotropy-aware layouts (Wang 2024).

Loss and Q.

α=2πngQiλ,αdB=4.343α(2.20)

At 1550 nm with ng = 2.3 template: Qi = 1e7 → 4.05 dB/m; 2.9e7 → 1.40 dB/m; 1.6e8 → 0.25 dB/m derived. Published conversions use each device's own ng (e.g. 1.3 dB/m for Qi 29 million reported).

Loss and Q records vs year
2016201820202022202420260.010.11101001000yearpropagation loss (dB/m)Zhang et al. 2017Desiatov et al. 2019, 637 nmLuke et al. 2020Gao et al. 2021Gao et al. 2022Zhu et al. 2024Qi 2.90e+07material limit 0.2 dB/m, Q 1.6e8bulk CLN absorption < 1e−4 /cm
■ loss reported□↓ Q only, ng 2.3 templatematerial limit reportedband derivedbulk CLN derivedconverter
2.300
1.550 µm
αdB  1.40 dB/mα 0.321 /m
Qi2.90e+07to floor  5.52×
Best dry-etched ring (Zhu et al. 2024) to the material floor: 5.5× in Q, 6.5× in loss derived
Type Qi (e.g. 1.6e8) or α; eq. 2.20 converts with the ng and λ set here. Qi slider 1e5 to 1e9.
0.1110100dB/mmaterial floor: Q 1.6e8, 0.2 dB/mbulk CLN absorption < 1e−4 /cm = 0.043 dB/m2020Luke272019Desiatov637 nm62017Zhang2.72024Zhu1.32021Gao>1e82022Gao>1e8Qi 2.90e+07
2.300
1.550 µm
Loss
  1.40 dB/m
α
 0.321 /m
Qi
2.90e+07
Floor Q / Qi
  5.52×

Best dry-etched ring to the floor: 5.5× in Q, 6.5× in loss. derived

Rim brightness = Q. Rings sorted by loss; dashed rims: Q only, placed with ng 2.3 at 1550 nm template. Band: material floor reported derived; dashed line: bulk absorption derived. Type 1.6e8 to land on the band.

Loss and Q records

PlatformLoss or QPlotted, dB/mDateSource
dry-etched TFLN ringQ about 1e7; 2.7 dB/m2.72017-12-18Zhang et al. 2017 reported
visible (637 nm)6 dB/m; Qi 11e662019-03-15Desiatov et al. 2019 reported
wafer-scale DUV lithography0.27 dB/cm27derived2020-08-03Luke et al. 2020 reported
microdiskQ > 1e8< 0.405template2021-12-01Gao et al. 2021 reported
CMP microringQ > 1e8< 0.405template2022Gao et al. 2022 reported
dry-etched ring, highest foundQi 29e6; 1.3 dB/m1.32024-07-26Zhu et al. 2024 reported
material limit, annealed ion-sliced filmQ 1.6e8; 0.2 dB/m0.2 reported2022-08-01Shams-Ansari et al. 2022 reported
bulk CLN absorption, 390 to 3800 nm< 1e−4 to 2 /cm0.043 derived2015Leidinger et al. 2015 reported

1e−4 /cm = 0.043 dB/m. Q-only rows are plotted as upper bounds on loss at ng = 2.3, 1550 nm.

L3 Mechanisms, computation, measurement

loss mechanisms, FEM, ring fits

Loss mechanisms

  • Sidewall and top-surface roughness scattering, set by lithography and Ar+ milling (redeposition: chapter 05).
  • Material absorption: transition-metal impurities give up to 0.04 /cm in as-grown crystals, H-related bands about 0.001 /cm at 2300 to 2800 nm; annealing lowers absorption up to 10× P. Ion-slicing damage: post-fabrication annealing raises the material-limited Q (Shams-Ansari et al. 2022).
  • Cladding absorption in the mid-IR: silica absorbs beyond 2.5 µm, which motivates LN on sapphire reported.
  • Bend radiation and substrate leakage (slab modes, thin buried oxide).
  • Metal absorption where electrodes approach the mode; Li et al. 2026 attribute their 2.8 dB on-chip loss to splitters, propagation and electrode absorption reported.
  • Photorefractive and charge effects at high power (chapter 04); removing the dielectric cladding mitigates photorefraction in rings reported.

Computing modes

Full-vector finite-element method (FEM) with the anisotropic permittivity rotated into the lab frame (x lateral, y vertical, z propagation; x-cut with propagation along crystal y puts crystal z along lab x) (eo-modulator-atlas engine). A scalar or effective-index solver is wrong for high-contrast anisotropic ribs. The neff 1.7696 and ng 2.135 engine of the Li et al. 2026 rib come from this solver.

Measuring loss

Ring transmission fits give loaded Q and, with the coupling regime identified (under- vs overcoupled), intrinsic Q; conversion to dB/m needs ng from the measured FSR:

FSR=λ2ngLrt

Cut-back measurements separate propagation from facet loss but need many lengths for sub-dB/cm accuracy.

Errors

Bulk Zelmon indices at 21 °C stand in for film indices; ion-sliced and annealed films may differ (composition, residual strain, Li out-diffusion: chapters 01, 05). Film-thickness nonuniformity across a wafer moves neff and therefore the QPM wavelength (2.3) and ring resonances. Real x-cut ring modes need an azimuthal treatment of the confined mode; the ring index above is the bulk indicatrix projection.

References

19 sources cited in this section
Desiatov et al. 2019Ultra-low-loss integrated visible photonics using thin-film lithium niobate. Optica 6 380 (2019).doi:10.1364/OPTICA.6.000380
Gao et al. 2021Broadband highly efficient nonlinear optical processes in on-chip integrated lithium niobate microdisk resonators of Q-factor above 10^8. New Journal of Physics 23 123027 (2021).doi:10.1088/1367-2630/ac3d52
Gao et al. 2022Lithium niobate microring with ultra-high Q factor above 10^8. Chinese Optics Letters 20 011902 (2022).doi:10.3788/col202220.011902
J. R. M.-L. Schwesyg 2011Interaction of light with impurities in lithium niobate crystals. Dr. rer. nat. dissertation, Universitat Bonn (2011).no DOI
Jiang 2026eo-modulator-atlas: FEM engine (electrostatics, full-vector optical modes, Pockels overlap). (2026).https://github.com/jwt625/eo-modulator-atlas
Jinan Jingzheng Electronics (NanoLN)Products.https://www.nanoln.com/PRODUCTS.html
Kaushalram et al. 2020Mode hybridization analysis in thin film lithium niobate strip multimode waveguides. Scientific Reports 10 16692 (2020).doi:10.1038/s41598-020-73936-x
Leidinger et al. 2015Optics Express 23 21690 (2015).doi:10.1364/OE.23.021690
Li et al. 2026Ultra-broadband near- to mid-infrared electro-optic modulator on thin-film lithium niobate. Nature Communications 17 1138 (2026).doi:10.1038/s41467-025-67902-2
Luke et al. 2020Wafer-scale low-loss lithium niobate photonic integrated circuits. Optics Express 28 24452 (2020).doi:10.1364/oe.401959
Mishra et al. 2021Mid-infrared nonlinear optics in thin-film lithium niobate on sapphire. Optica 8 921 (2021).doi:10.1364/OPTICA.427428
Pan et al. 2019Fundamental mode hybridization in a thin film lithium niobate ridge waveguide. Optics Express 27 35659 (2019).doi:10.1364/oe.27.035659
Shams-Ansari et al. 2022Reduced material loss in thin-film lithium niobate waveguides. APL Photonics 7 081301 (2022).doi:10.1063/5.0095146
Tan et al. 2024Intrinsic polarization conversion and avoided-mode crossing in X-cut lithium niobate microrings. Chinese Physics B 33 064205 (2024).doi:10.1088/1674-1056/ad47b2
Wang 2024Breaking anisotropy limitations in thin-film lithium niobate arrayed waveguide gratings. Light: Science & Applications 13 209 (2024).doi:10.1038/s41377-024-01551-w
Xu et al. 2021Mitigating photorefractive effect in thin-film lithium niobate microring resonators. Optics Express 29 5497 (2021).doi:10.1364/OE.418877
Zelmon et al. 1997Infrared corrected Sellmeier coefficients for congruently grown lithium niobate and 5 mol% magnesium oxide –doped lithium niobate. Journal of the Optical Society of America B 14 3319 (1997).doi:10.1364/JOSAB.14.003319
Zhang et al. 2017Monolithic ultra-high-Q lithium niobate microring resonator. Optica 4 1536 (2017).doi:10.1364/OPTICA.4.001536
Zhu et al. 2024Twenty-nine million intrinsic Q-factor monolithic microresonators on thin-film lithium niobate. Photonics Research 12 A63 (2024).doi:10.1364/prj.521172