README.md

---
language: 
  - en
size_categories:
  - 50K<n<100K
license: mit
task_categories:
  - tabular-regression
tags:
  - photonics
  - silicon-nitride
  - waveguide
  - optical
  - dataset
  - synthetic
dataset_info:
  features:
    - name: waveguide_width
      dtype: float
    - name: waveguide_height
      dtype: float
    - name: cladding_material
      dtype: string
    - name: cladding_thickness
      dtype: float
    - name: deposition_method
      dtype: string
    - name: etch_method
      dtype: string
    - name: sidewall_roughness
      dtype: float
    - name: annealing_params
      dtype: string
    - name: wavelength
      dtype: float
    - name: polarization
      dtype: string
    - name: input_power
      dtype: float
    - name: temperature
      dtype: float
    - name: bend_radius
      dtype: float
    - name: device_length
      dtype: float
    - name: insertion_loss
      dtype: float
    - name: propagation_loss
      dtype: float
    - name: coupling_efficiency_input
      dtype: float
    - name: coupling_efficiency_output
      dtype: float
    - name: scattering_loss
      dtype: float
    - name: effective_index
      dtype: float
    - name: mode_confinement_factor
      dtype: float
    - name: batch_id
      dtype: string
    - name: data_source
      dtype: string
    - name: measurement_method
      dtype: string
    - name: uncertainty
      dtype: float
dataset_size: 90000
dataset_version: "1.0.0"
---
# SiN Photonic Waveguide Loss & Efficiency Dataset

> **Description**  
> This dataset provides **90,000 synthetic rows** of silicon nitride (Si₃N₄) photonic waveguide parameters, focusing on **waveguide loss** and **efficiency** metrics. The data is useful for modeling, simulation, or LLM fine tuning to predict and understand the relationship between fabrication/design parameters and optical performance.

## Key Highlights ✨
- **Material Focus**: Silicon Nitride (Si₃N₄)  
- **Columns**: 25 structured columns capturing waveguide geometry, fabrication method, operational conditions, and measured/synthetic performance metrics  
- **Size**: 90,000 rows (ideal for both training and validation splits)  
- **Use Cases**:  
  - Waveguide loss prediction  
  - Process control and optimization  
  - Photonic design parameter studies  
  - Synthetic data augmentation for AI/ML tasks  

## Dataset Structure 🏗️
Each row corresponds to a **single waveguide configuration** or measurement instance, including:

1. **Waveguide Geometry**  
   - `waveguide_width` (µm)  
   - `waveguide_height` (nm or µm)  
   - `bend_radius` (µm)  
   - `device_length` (mm)

2. **Material & Fabrication**  
   - `cladding_material`  
   - `cladding_thickness` (µm)  
   - `deposition_method`  
   - `etch_method`  
   - `sidewall_roughness` (nm)  
   - `annealing_params`

3. **Operational Parameters**  
   - `wavelength` (nm)  
   - `polarization` (TE/TM)  
   - `input_power` (dBm)  
   - `temperature` (°C)

4. **Performance Metrics**  
   - `insertion_loss` (dB)  
   - `propagation_loss` (dB/cm)  
   - `coupling_efficiency_input` (%)  
   - `coupling_efficiency_output` (%)  
   - `scattering_loss` (dB/cm)  
   - `effective_index`  
   - `mode_confinement_factor` (0–1)

5. **Metadata**  
   - `batch_id` (fabrication batch/wafer ID)  
   - `data_source` (Synthetic or Measurement)  
   - `measurement_method` (e.g., cut-back, ring_resonance)  
   - `uncertainty` (± dB or %)

## Example Row
waveguide_width                = 1.212  
waveguide_height               = 400.00  
cladding_material              = SiO2  
cladding_thickness             = 2.50  
deposition_method              = LPCVD  
etch_method                    = RIE  
sidewall_roughness             = 2.05  
annealing_params               = 900C_3hr  
wavelength                     = 1552.23  
polarization                   = TE  
input_power                    = 0.00  
temperature                    = 25.00  
bend_radius                    = 50.00  
device_length                  = 10.00  
insertion_loss                 = 3.50  
propagation_loss               = 0.300  
coupling_efficiency_input      = 72.00  
coupling_efficiency_output     = 68.00  
scattering_loss                = 0.15  
effective_index                = 1.800  
mode_confinement_factor        = 0.80  
batch_id                       = BATCH_12  
data_source                    = Synthetic  
measurement_method             = ring_resonance  
uncertainty                    = 0.05  

## How to Use 💡
1. **Download/Clone**  
   - You can download the CSV file manually or use Hugging Face’s `datasets` library:  
     ```python
     from datasets import load_dataset

     dataset = load_dataset("username/SiN_Photonic_Waveguide_Loss_Efficiency")
     ```

2. **Loading & Exploration**  
   - Load into your favorite Python environment (`pandas`, `polars`, etc.) to quickly explore the data distribution:
     ```python
     import pandas as pd

     df = pd.read_csv("SiN_Photonic_Waveguide_Loss_Efficiency.csv")
     print(df.head())
     ```

3. **Model Training**  
   - For tasks like waveguide loss prediction, treat the waveguide geometry/fabrication columns as input features, and the `insertion_loss` or `propagation_loss` columns as the labels or targets.  
   - Example ML scenario:  
     ```python
     features = df[[
         "waveguide_width", "waveguide_height", "sidewall_roughness", 
         "wavelength", "polarization", "temperature"
     ]]
     target = df["propagation_loss"]

     # Then train a regression model, e.g., scikit-learn, XGBoost, etc.
     ```

4. **Synthetic Data Augmentation**  
   - Use this synthetic dataset to **supplement** smaller real datasets, enabling data-hungry deep learning models to generalize better.

## Dataset Creation Process 🛠️
A Python script was used to randomly generate each column’s values within plausible ranges based on typical Si₃N₄ waveguide fabrication and performance data. The insertion loss is partially derived from the propagation loss and device length, and additional random offsets account for coupling losses and measurement variability.

## Caveats & Limitations ⚠️
- **Synthetic Nature**: While ranges are inspired by real-world photonic designs, actual values may differ based on specific foundries, tools, and processes.  
- **Statistical Simplifications**: Not all real-world correlations or distributions (e.g., non-uniform doping profiles, advanced thermal effects) are captured.  
- **Measurement Noise**: The `uncertainty` column does not fully replicate complex measurement artifacts.  

## License 📄
This dataset is available under the **MIT License**. You are free to modify, distribute, and use it for commercial or non-commercial purposes—just provide attribution.

## Citation & Acknowledgments 🙌
If you use this dataset in your research or applications, please cite it as follows (example citation):

> **Author**: _https://huggingface.co/Taylor658_  
> **Title**: _SiN Photonic Waveguide Loss & Efficiency (Synthetic)_  
> **Year**: 2025

```bibtex
@misc{sin_waveguide_loss_efficiency_2025,
  title  = {SiN Photonic Waveguide Loss & Efficiency (Synthetic)},
  author = {atayloraeropsace},
  year   = {2025},
  howpublished = {\url{https://huggingface.co/datasets/username/SiN_Photonic_Waveguide_Loss_Efficiency}}
}
```

## Contributing 🧑‍💻
We welcome community contributions, ideas, and corrections:
- **Add additional columns** (e.g., doping profiles, stress levels, advanced measurement data).  


---