jjquaratiello commited on
Commit
63f98e4
·
verified ·
1 Parent(s): a9738e0

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +113 -0
README.md ADDED
@@ -0,0 +1,113 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ tags:
4
+ - physics
5
+ - electromagnetic-simulation
6
+ - photonics
7
+ - silicon-photonics
8
+ - fdtd
9
+ - pde
10
+ size_categories:
11
+ - 10K<n<100K
12
+ ---
13
+
14
+ # PIC-Flow Dataset
15
+
16
+ 22,500 frequency-domain FDTD electromagnetic-field simulations for parameterized
17
+ silicon-on-insulator photonic devices at λ = 1.55 µm. Used as the training,
18
+ validation, and test data for the [PIC-Flow](https://huggingface.co/RizzoLab/PIC-Flow)
19
+ neural surrogate model.
20
+
21
+ Code, documentation, and inference notebooks live in the GitHub repo:
22
+ [Rizzo-Integrated-Photonic-Systems-Lab/PIC-Flow](https://github.com/Rizzo-Integrated-Photonic-Systems-Lab/PIC-Flow).
23
+
24
+ ## Contents
25
+
26
+ | Path | Description |
27
+ |---|---|
28
+ | `shards/shard_*.npz` | Packed FDTD samples (~225 shards). Each shard contains many slots `s0/`, `s1/`, ... |
29
+ | `shards/index.json` | Manifest mapping `(device, geometry_id, split, augment) → (shard, slot)`. |
30
+ | `geometries.jsonl` | One line per simulated device: family, geometric parameters, port count. |
31
+
32
+ ## Per-sample fields
33
+
34
+ Each slot inside a shard carries:
35
+
36
+ | Key | Shape | Description |
37
+ |---|---|---|
38
+ | `eps` | `(160, 480)` float32 | Relative permittivity ε_r (Si core ≈ 5.8, SiO₂ cladding ≈ 2.09). |
39
+ | `Ez_real`, `Ez_imag` | `(160, 480)` float32 | Real and imaginary parts of the complex E_z field, source-anchored phase. |
40
+ | `src_mask` | `(160, 480)` float32 | Binary mask marking the active eigenmode-source port. |
41
+ | `port_masks` | `(N_ports, 160, 480)` float32 | Per-port binary masks (e.g., 4 ports for MMIs/DCs, 3 for Y-branches). |
42
+ | `port_ids` | `(N_ports,)` int32 | Integer port labels matching `port_masks`. |
43
+ | `input_port` | int | Which port was excited by the eigenmode source. |
44
+ | `wavelength_um` | float | Free-space wavelength in µm (1.55 throughout this dataset). |
45
+ | `dx_um`, `dy_um`, `Lx_um`, `Ly_um` | float | Grid resolution and physical extent. |
46
+ | `device` | string | Family: `mmi`, `ybranch`, or `directional_coupler`. |
47
+ | `geometry_id`, `split` | string | Unique geometry id and `train` / `val` / `test` membership. |
48
+ | `params/<name>` | float | Geometric parameters for this device (varies by family — see below). |
49
+
50
+ ## Devices and parameter sweep
51
+
52
+ 5-dimensional Latin-hypercube sweep per family, quantized to a half-pixel grid
53
+ (0.025 µm at 20 pixels/µm):
54
+
55
+ - **2×2 MMI** (4 ports, 7,500 samples): waveguide width [0.40–0.575], MMI width [4.5–5.5],
56
+ MMI length [8.0–15.0], taper width [0.575–1.5], taper length [1.0–3.0] µm.
57
+ - **Y-branch** (3 ports, 7,500): waveguide width [0.40–0.575], junction length [1.0–3.0],
58
+ bend length [4.0–7.0], arm offset [0.575–2.5], output length [1.0–4.0] µm.
59
+ - **Directional coupler** (4 ports, 7,500): waveguide width [0.40–0.575], gap [0.10–0.35],
60
+ coupling length [5.0–8.0], bend length [4.0–6.0], port separation [0.825–2.0] µm.
61
+
62
+ ## Splits
63
+
64
+ Index-based, shared across all PIC-Flow ablation runs:
65
+
66
+ | Split | Samples |
67
+ |---|---|
68
+ | train | 18,000 |
69
+ | val | 2,250 |
70
+ | test | 2,250 |
71
+
72
+ ## How to download
73
+
74
+ ```bash
75
+ pip install huggingface_hub
76
+ hf download RizzoLab/PIC-Flow-Dataset --repo-type dataset --local-dir Data/unified_sweep_mmi_ybranch_dc_7500_each_1p55um
77
+ ```
78
+
79
+ The dataloader in the GitHub repo
80
+ ([`Model/dataset.py`](https://github.com/Rizzo-Integrated-Photonic-Systems-Lab/PIC-Flow/blob/main/Model/dataset.py))
81
+ expects this layout under `Data/unified_sweep_mmi_ybranch_dc_7500_each_1p55um/`.
82
+
83
+ ## Generation
84
+
85
+ The data was produced with the Meep FDTD solver via
86
+ [`FDTD/unified_sweep.py`](https://github.com/Rizzo-Integrated-Photonic-Systems-Lab/PIC-Flow/blob/main/FDTD/unified_sweep.py)
87
+ in the GitHub repo. Each sample uses an eigenmode source exciting the fundamental TE
88
+ mode at the selected input port; fields are extracted at λ = 1.55 µm via discrete
89
+ Fourier transforms. The vertical 220 nm SOI slab mode is pre-solved into an effective
90
+ core index n_eff ≈ 2.4 used for the 2D scalar Helmholtz simulation.
91
+
92
+ To regenerate from scratch (~24 hours on one CPU node, 16 threads):
93
+
94
+ ```bash
95
+ python FDTD/unified_sweep.py --output-dir Data/ \
96
+ --devices mmi,ybranch,directional_coupler \
97
+ --num-samples 7500 --wavelengths 1.55
98
+ ```
99
+
100
+ ## Citation
101
+
102
+ ```bibtex
103
+ @article{Quaratiello2026PICFlow,
104
+ author = {Joseph Quaratiello and Anthony Rizzo},
105
+ title = {Physics-Based Flow Matching for Full-Field Prediction of Silicon Photonic Devices},
106
+ journal = {arXiv},
107
+ year = {2026}
108
+ }
109
+ ```
110
+
111
+ ## License
112
+
113
+ MIT.