Pixel Structure Realism Report

Generated on 2026-06-06 from compass.geometry.sample_pixels and PixelStack.

This report documents the structural-realism features COMPASS models for modern backside-illuminated (BSI) CMOS image-sensor pixels, and visualises them directly from the generated solver input. The cross-sections sample the real permittivity Re(eps)(x, z) the RCWA solver integrates, so this is geometry evidence rather than an optical-performance claim.

Why these features matter

• Backside inverted-pyramid texture (IPA). A graded silicon fill fraction at the backside acts as a moth-eye anti-reflection / light-trapping layer. Published simulations and process reports show large near-infrared QE gains (up to ~3x at 850 nm and ~5x at 940 nm) when an IPA is combined with deep DTI.
• DTI conformal high-k liner. Real backside DTI trenches are lined with a thin high-k film (Al2O3 / HfO2 / Ta2O5, ~30-100 nm) that passivates the etched silicon and carries a negative fixed charge. Optically it is a thin high-index ring between silicon and the lower-index trench fill.
• Tapered DTI sidewall. Etched trenches narrow with depth; a vertical-wall idealisation over-counts the isolation oxide deep in the substrate.
• Microlens residual base. Reflow / etch-back leaves a flat polymer slab under the curved cap; the lens is never zero-thickness at its edges.

Baseline vs NIR-enhanced pixel

Silicon backside detail

The graded Re(eps) of the inverted-pyramid texture produces a smooth effective-index transition from the trench fill toward bulk silicon, while the tapered DTI trench and its high-k liner are visible at the pixel boundary.

• Texture height: 0.350 um
• Area-averaged effective index across the texture: 1.46 at the surface to 3.67 toward the apex (monotonic graded-index anti-reflection).

Capability coverage matrix

Real-pixel feature	COMPASS status	config key
Microlens superellipse profile + CRA shift	modelled	microlens.profile, microlens.shift
Multi-pixel shared lens (2x2 / 4x4 OCL)	modelled	microlens.sharing
Microlens residual base layer	modelled (new)	microlens.base_thickness
Per-color CF thickness + contact-angle relief	modelled	color_filter.<color>.thickness/contact_angle
Metal grid (W) with rounded corners	modelled	color_filter.grid
BARL anti-reflection multilayer	modelled	barl.layers
FDTI / BDTI deep trench isolation	modelled	silicon.dti.mode/depth/width
DTI conformal high-k passivation liner	modelled (new)	silicon.dti.liner
Tapered DTI sidewall (etch profile)	modelled (new)	silicon.dti.taper_angle
Backside inverted-pyramid light-trapping texture	modelled (new)	silicon.surface_texture
Photodiode collection window	modelled	silicon.photodiode
Composite / air-gap metal grid liner	roadmap	-
In-pixel light pipe / inner lens	roadmap	-

Reproduce this pixel

python scripts/run_simulation.py pixel=sample_p1p12um_nir

Or derive a config in Python:

from compass.geometry.sample_pixels import derive_parameters
from compass.geometry.pixel_stack import PixelStack

cfg = derive_parameters("sample_p1p12um_nir")
stack = PixelStack({"pixel": cfg})

Regeneration

python scripts/generate_realism_report.py

Generated metrics are stored at docs/public/reports/geometry/structure-realism/structure_realism_metrics.json.