Anthracene in-plane cavity

In-plane Si3N4: TIR-tunnel excitation and an inverse-designed + fishbone-reflector Purcell cavity.

In-plane Si3N4 structures for:

1. TIR tunnel excitation — angled pump WG → Ag film → 200 nm crystal → collection WG
2. Fishbone cavity — invdes + crystal + collection WG + fishbone reflector (Purcell + η)

Dry-run only (default)

cd anthracene_inplane_cavity_v1
python run_dry_run.py

• Builds and validates all planned simulations
• High-contrast structure maps (monitors/sources hidden) under figures/s1_tir/ and figures/s2_cavity/
• Quotes FlexCredits via Job.estimate_cost (not billed)
• Two reports:
• reports/s1_tir_dry_run_report.html — pump → Ag → crystal → collection
• reports/s2_invdes_cavity_dry_run_report.html — invdes + fishbone cavity

Do not run FDTD until you explicitly approve FlexCredit spend.

Layout

Path	Role
anthracene_inplane_cavity_v1/config.py	Materials, sweeps, fishbone params
geometry_s1_tir.py	Structure 1 coplanar TIR geometry
geometry_s2_cavity.py	Structure 2 cavity variants
fishbone_port.py	Wrapper → ../fishbone_reflector/fishbone_geometry.py
metrics.py	Metric defs + analytic Ag/TIR precheck
run_dry_run.py	Dry-run entry point

Decisions

• Core: Si3N4 n=2.0, t=300 nm
• Crystal: 200 nm anthracene
• Coplanar waveguides (Structure 1)
• λ = 780 nm

Results

Stage 1 -- first simulations (baseline)

S1 first simulations: dipole collection + pump excitation

Estimated FC (sum): 0.8084 · Actual FC (sum): 0.1450

Job	Estimated FC	Actual FC	Task
s1_first_dipole	0.4018	0.0719	fdve-1b06d96d-5137-4119-b496-78ca7b716762
s1_first_pump	0.4066	0.0730	fdve-6dd2f296-0983-41ea-a327-2c1179c8bc7d
Total	0.8084	0.1450

Geometry: θ = 55°, d_Ag = 60 nm, crystal = 200 nm in x (∞ yz), collection butted to crystal, λ = 780 nm, Si₃N₄ strip 400×300 nm.

Dipole emission → waveguides

Quantity	Value
P_total (signed outward power-box flux)	1764
Flux through collection plane	789.2
Flux through pump plane	-22.81
**η_coll =	Φ_coll
**η_pump =	Φ_pump
η_coll mode-0 only (ModeMonitor)	41.6387%
η_coll all modes (ModeMonitor)	41.6387%
η_pump mode-0 only (ModeMonitor)	0.7581%
η_pump all modes (ModeMonitor)	0.7581%

Primary metrics are flux-based. Field maps show a clear guided beam in the collection waveguide with Ey-dominant polarization matching the Ey dipole. ModeMonitor |a|² is orders of magnitude below the collection FluxMonitor for this job (likely continuum contamination / mode-expansion failure on the large plane) — listed above for transparency but not used as the headline η.

E-field — dipole on (coupling into collection WG)

Dipole |E| x–y — guided collection

Dipole log10|E| x–y

Dipole |E| x–z

Dipole |E| y–z in crystal

Pump on — excitation / evanescent field / cross-talk

Quantity	Value
P_in (pump mode-0 + direction)	0.3171
P_collection mode-0 from pump	0.0007105
Cross-talk X_mode = P_coll / P_in	2.240e-03
X_mode (dB)	-26.5 dB
	Φ_pump

Linear |E| shows strong confinement in the angled pump guide and reflection at the Ag face. Log-scale and the x-linecut show residual field tunneling through Ag into the crystal (evanescent / thin-metal transmission) and weak scatter into the collection arm.

E-field — pump on (TIR / tunnel / evanescent)

Pump |E| x–y — TIR at Ag

Pump log10|E| x–y — tunnel / scatter

Pump |E|(x) line cut through Ag/crystal

Pump |E| x–z

Pump |E| y–z in crystal (emitter plane)

Notes

• η_coll / η_pump (headline): |FluxMonitor at WG plane| / total outward power-box flux. This counts all power crossing that plane (guided + co-propagating radiation). Field maps indicate most collection-plane power is in the strip-guided beam.
• ModeMonitor caveat: for the dipole job, mode-0 |a|² ≪ flux; do not trust mode-only η here until monitors are retuned (tighter mode plane, more modes, or offline ModeSolver projection).
• Cross-talk X uses ModeSource fundamental power (pump job); same-λ worst case. A real Stokes-shifted pump + filter would suppress detector contamination further.
• Early shutoff ~10% of run_time (field decay < 1e-5); guided path lengths are short enough that this is OK.

Stage 7 -- DBR + spacer inverse design (final stage)

S7 DBR + SiO₂ spacer invdes — spacer thickness sweep (10 steps each)

Stack: SiO₂ substrate (→−z PML) → 8-pair TiO₂/SiO₂ DBR → SiO₂ spacer (swept) → TiO₂ invdes 4×4 µm × 250 nm → crystal → Si₃N₄ WG ‖ z. No thin TiO₂ mirror film.

Best by final η: SiO₂ spacer = 200.0 nm → η ≈ 82.20%

Hyperparameters (shared; SiO₂ spacer varies)

Parameter	Value
Adam steps	10
Learning rate	0.25
Filter β	8.0
Filter radius	80 nm
Pixel size	40 nm
Init fill	0.5
Penalty weight / length	0.5 / 80 nm
Objective	η = P_mode / P_box
DBR pairs	8 TiO₂/SiO₂ QW @ 780 nm
QW TiO₂ / SiO₂	69.6 / 134.5 nm
SiO₂ spacer	200 nm (swept)
Thin TiO₂ film	none
Invdes	4×4 µm × 250 nm SiO₂↔TiO₂
Crystal / WG	100 nm ∞xy / Si₃N₄ 450×300 nm
Symmetry	(-1, +1, 0)

QW thicknesses @ 780 nm: TiO₂ 69.6 nm, SiO₂ 134.5 nm. Spacer sweep: ['100', '200', '300'] nm.

Comparison (which performs best)

SiO₂ spacer (nm)	Best η (opt)	Final η (re-eval)	P_mode	F_p@780	F_p max	FC
100.0	79.81%	79.84%	2.184e+04	1.480	1.695 @ 765 nm	0.5250
150.0	79.93%	80.11%	1.205e+04	0.814	2.432 @ 845 nm	0.5250
200.0	81.92%	82.20%	1.102e+04	0.725	2.047 @ 855 nm	0.5250
250.0	81.64%	81.89%	2.954e+04	1.953	1.954 @ 775 nm	0.5250
300.0	62.92%	66.55%	1.171e+04	0.950	1.087 @ 730 nm	0.5250

Purcell spectra

All three spacer thicknesses

SiO₂ spacer = 100.0 nm (sp100)

η: best 79.81%, final re-eval 79.84% · F_p(780)=1.480

E-field profiles (required)

sp100 |E| xy

sp100 |E| xz

Structure / history / Purcell

sp100 density

sp100 xz

sp100 history

sp100 Purcell

sp100 init xz (DBR visible)

SiO₂ spacer = 150.0 nm (sp150)

η: best 79.93%, final re-eval 80.11% · F_p(780)=0.814

E-field profiles (required)

sp150 |E| xy

sp150 |E| xz

Structure / history / Purcell

sp150 density

sp150 xz

sp150 history

sp150 Purcell

sp150 init xz (DBR visible)

SiO₂ spacer = 200.0 nm (sp200)

η: best 81.92%, final re-eval 82.20% · F_p(780)=0.725

E-field profiles (required)

sp200 |E| xy

sp200 |E| xz

Structure / history / Purcell

sp200 density

sp200 xz

sp200 history

sp200 Purcell

sp200 init xz (DBR visible)

SiO₂ spacer = 250.0 nm (sp250)

η: best 81.64%, final re-eval 81.89% · F_p(780)=1.953

E-field profiles (required)

sp250 |E| xy

sp250 |E| xz

Structure / history / Purcell

sp250 density

sp250 xz

sp250 history

sp250 Purcell

sp250 init xz (DBR visible)

SiO₂ spacer = 300.0 nm (sp300)

η: best 62.92%, final re-eval 66.55% · F_p(780)=0.950

E-field profiles (required)

sp300 |E| xy

sp300 |E| xz

Structure / history / Purcell

sp300 density

sp300 xz

sp300 history

sp300 Purcell

sp300 init xz (DBR visible)

29 additional stage reports (s1-s7, dry runs and intermediate sweeps) exist under reports/.

Downloads

FDTD simulation study only -- no GDS/STL deliverable files exist in this tree (the fishbone reflector mirror it uses has its own GDS -- see the Fishbone reflector project).

---

Source: nanophotonic_devices/cavity_purcell/anthracene_inplane_cavity_v1/