#4846 Etched Glass Stack
#4846 Etched Glass Stack
#4846 Concept: Stack of etched glass lenses to direct light in a desired pattern for a desired purpose
A stack of planar, etched glass elements functioning in series operates by breaking continuous wave propagation into discrete phase-delay transformations. Rather than relying on bulky curved bulk optics, each layer introduces spatially localized changes in phase, amplitude, or optical path length, allowing incoming wavefronts to be steered, split, or computed upon at the speed of light.
Core Physical Mechanisms
When chaining etched glass layers, three distinct physical regimes govern how light is routed:
Multi-Plane Diffractive Optical Networks (Free-Space Cascades):
Each layer consists of sub-wavelength or multi-level phase steps etched into fused silica or borosilicate glass.
As light passes through Layer N, it acquires a spatially varying phase shift Delta phi(x, y) = {2\pi} / {lambda} (n_glass - n_air) d(x, y), where d(x, y) is the local etch depth.
Between plates, light undergoes angular spectrum / Rayleigh-Sommerfeld diffraction through the air gap, acting as a global interconnect where every point on Layer N diffracts toward multiple points on Layer N+1.
Cascading these layers implements physical, all-optical linear transformations (analogous to deep neural network weight layers or multi-stage beam shapers) without needing electro-optic conversion.
Evanescent & Inter-Layer Waveguide Coupling (Guided-Wave Stacks):
If the glass plates feature surface relief or subsurface laser-written waveguide channels and are stacked with index-matched optical adhesive (or direct molecular bonding), light couples across layers via evanescent fields.
Directional couplers, 3D routing manifolds, and multi-mode interference (MMI) splitters can shift beams up, down, or laterally across the 3D volume.
Multi-Layer Metasurface / Flat Lens Stacks:
Etched sub-wavelength nanopillars or high-contrast gratings impart abrupt phase discontinuities.
Stacking multiple planar metasurface plates compensates for chromatic aberration and wide-angle off-axis coma, yielding high numerical aperture (NA) telecentric lens assemblies within millimeters of track length.
Key Applications & Architectural Classes
| Architecture | Operational Principle | Primary Application / Output |
| Diffractive Deep Neural Networks ($D^2NN$) | Cascaded phase plates implementing complex matrix-vector products through free-space diffraction. | All-optical object classification, spatial filtering, and high-throughput analog inference at near-zero operating power. |
| Structured Light & Holographic Projectors | Segmented phase modulation distributing a Gaussian or flat-top beam into arbitrary 2D/3D topologies. | Dot-matrix LiDAR projectors, 3D surface profilometry, and material processing beam-homogenizers. |
| 3D Optical Cross-Connects | Waveguide-inscribed glass wafers linked via directional coupling or vertical turn mirrors. | High-density photonic routing, multi-core fiber fan-outs, and chip-to-chip optical backplanes. |
| Ultra-Compact Corrective Imaging | Alternating positive/negative phase profiles etched across flat glass plates. | Wafer-level micro-cameras, endoscopes, and head-mounted display (AR/VR) beam expanders. |
Fabrication & Implementation Pathways
Surface Photolithography & Reactive Ion Etching (RIE): Suitable for discrete binary or multi-level phase plates (e.g., 8-level or 16-level phase profiles). Alignment fiducials and kinematic mounts (or spacers with UV-curable optical resin) are required to control the inter-layer distance (z) to sub-micron tolerances.
Femtosecond Laser Direct Writing (FLDW): Instead of physically etching surface features and stacking separate plates, an ultrafast laser modifies the internal refractive index (Delta n sim 10^-3 to 10^-2) inside a monolithic fused silica block. This produces true 3D embedded waveguides without the reflection losses or mechanical drift inherent to loose multi-slide stacks.
Etch-and-Bond (Direct Wafer Bonding): Slides etched with microchannels or relief patterns are polished to sub-nanometer roughness and bonded via oxygen plasma activation or anodic bonding, forming a solid, monolithic glass engine immune to environmental misalignment.
Concrete Implementation Checklist
Calculate the Rayleigh Range & Layer Spacing: Determine spacing z between plates based on the minimum feature pitch p and design wavelength lambda; ensure z sim p^2 / lambda so high-frequency spatial modes properly interfere prior to the next plate.
Phase Discretization: For etched phase steps, target 2^N depth levels (N = 3 or 4) using depth formula d = {m lambda} / {2^N (n - 1)} to keep diffraction efficiency above 95%.
Anti-Reflective (AR) Interfacing: If maintaining air gaps between slides, deposit broadband dielectric AR coatings on unetched flat faces to eliminate etalon fringes and multi-path Fabry-Pérot reflections between adjacent plates.
Substrate Selection: Use fused silica (n ~ 1.458 at 589 nm) for thermal stability and high optical damage threshold under high-power laser illumination, or high-index optical glasses (e.g., N-SF11) when tighter phase delays and steeper numerical apertures are required.
Describe how setup could be simulated at the macro level
Simulating a macro-level stack of etched glass phase plates—often referred to as a Diffractive Optical Network or Multi-Plane Light Conversion (MPLC) system—involves treating each etched glass plate as a thin phase screen separated by free-space wave propagation.
At the macro scale (where spatial features are typically larger than the operating wavelength, p > lambda), rigorous Maxwell equation solvers like FDTD (Finite-Difference Time-Domain) are computationally prohibitive. Instead, the standard macroscopic framework is the Angular Spectrum Method (ASM) or Split-Step Fourier Method (SSFM).
1. Mathematical Simulation Model
The optical field is represented on a 2D discrete grid as a complex scalar electric field:
Propagation through an N-plate stack alternates between two operations:
[Incident Beam U_in] │ ▼ ┌──────────────┐ │ Phase Plate 1│ ──> Multiply by T_1(x, y) = exp(i * Δφ_1(x, y)) └──────────────┘ │ ▼ ┌──────────────┐ │ Free Space │ ──> Angular Spectrum Propagation H(u, v; Δz) └──────────────┘ │ ▼ ┌──────────────┐ │ Phase Plate 2│ ──> Multiply by T_2(x, y) = exp(i * Δφ_2(x, y)) └──────────────┘ │ ... ▼ [Target Intensity / Field Profile]
Step A: Thin Phase Screen Modulation
Passing through Plate k, the local etch depth h_k(x, y) introduces a spatially dependent optical path delay:
Step B: Inter-Plate Angular Spectrum Propagation
Between plates separated by axial distance Delta z, the field propagates via the Fourier transfer function H(u, v):
(Evanescent modes where (lambda u)^2 + (lambda v)^2 > 1 are set to zero).
2. Grid Sizing & Anti-Aliasing Constraints
To avoid spatial aliasing and numerical wraparound artifacts, the simulation parameters must satisfy sampling criteria:
| Parameter | Constraint | Operational Reason |
| Sampling Pitch (Delta x, \Delta y) | Delta x <= {lambda Delta z} / {L} | Ensures highest spatial frequencies are resolved without phase folding across plate gap Delta z. |
| Grid Window (L = N Delta x) | L >= D_beam + 2 \Delta z tan(theta_max) | Prevents diffracted light from hitting boundaries and wrapping around (circular convolution artifact). |
| Absorbing Boundary Layer | Outer 5–10% padded with a Hann/Tukey taper | Dampens stray high-angle scatter instead of reflecting it back into the active field. |
3. Macro Optical Simulator (Python Reference)
This self-contained Python routine sets up an incident Gaussian beam, routes it through a 3-layer etched phase plate stack using ASM, and plots the resulting phase and intensity profiles across the optical axis.
import numpy as np
import matplotlib.pyplot as plt
def simulate_etched_glass_stack():
# --- Physical Parameters ---
wavelength = 632.8e-9 # 632.8 nm (HeNe Laser)
n_glass = 1.515 # Borosilicate / BK7 refractive index
k = 2 * np.pi / wavelength
N = 1024 # Grid resolution
L = 5.0e-3 # Simulation domain: 5 mm x 5 mm
dx = L / N
x = np.linspace(-L/2, L/2, N)
y = np.linspace(-L/2, L/2, N)
X, Y = np.meshgrid(x, y)
# --- Spatial Frequency Coordinates for ASM ---
fx = np.fft.fftfreq(N, d=dx)
fy = np.fft.fftfreq(N, d=dx)
FX, FY = np.meshgrid(fx, fy)
# Angular spectrum transfer function kernel generator
def propagate(U, z):
arg = 1.0 - (wavelength * FX)**2 - (wavelength * FY)**2
valid = arg >= 0
H = np.zeros((N, N), dtype=complex)
H[valid] = np.exp(1j * k * z * np.sqrt(arg[valid]))
return np.fft.ifft2(np.fft.fft2(U) * H)
# --- Initial Field: Collimated Gaussian Beam ---
w0 = 0.8e-3 # 0.8 mm waist
U = np.exp(-(X**2 + Y**2) / (w0**2)).astype(complex)
# --- Define 3 Etched Plates & Distances ---
dz = 15.0e-3 # 15 mm air gap between successive plates
# Plate 1: Etched Cylindrical Fresnel lens (focuses x-axis)
fx1 = 30e-3
etch_1 = ((X**2) / (2 * fx1)) % (wavelength / (n_glass - 1))
phi_1 = k * (n_glass - 1) * etch_1
# Plate 2: Etched Beam Splitter Grating (splits into ±1 orders)
grating_pitch = 120e-6
etch_2 = (np.sin(2 * np.pi * Y / grating_pitch) > 0) * (wavelength / (2 * (n_glass - 1)))
phi_2 = k * (n_glass - 1) * etch_2
# Plate 3: Annular / Vortex Phase Step (adds helical orbital angular momentum)
theta = np.arctan2(Y, X)
phi_3 = 2 * theta # Topological charge m = 2
plates = [phi_1, phi_2, phi_3]
# --- Forward Cascade ---
fields_after_plates = []
for idx, phi in enumerate(plates):
# 1. Apply plate phase modulation
U = U * np.exp(1j * phi)
# 2. Propagate across inter-plate gap
U = propagate(U, dz)
fields_after_plates.append(U)
# Output target intensity
target_intensity = np.abs(fields_after_plates[-1])**2
return X, Y, target_intensity
# Run simulation
X, Y, intensity = simulate_etched_glass_stack()
4. Accounting for Physical Non-Idealities
To match bench measurements, macro simulations add four error models to the ideal transfer function:
Substrate Bulk Transmission Delay: Glass plates of thickness t_g add a constant phase offset exp(i k n_glass t_g) and mild beam expansion inside the glass.
Surface Reflection & Fabry-Pérot Interference: At each glass-air interface, Fresnel reflections (R ~ left({n-1} / {n+1} right)^2 ~ 4%) produce ghost orders. This can be incorporated via transfer matrix methods or simple attenuation scaling:
T_interface = 1 - R ~ 0.96Etch Depth Discretization (Quantization Noise): If lithographic etching uses $M discrete mask levels (e.g., 8 or 16 phase steps), continuous depths are rounded to:
h_{quantized = round} left(h cdot {2^M - 1 / }{h_max} right) {h_max} / {2^M - 1}Alignment Drift: Add rigid body translation (\delta x, delta y) and tilt (theta_x, theta_y) kernels to U_k(x, y) between layers to assess tolerance bounds before sending masks to fabrication.
Concrete Next Steps
Fix Wavelength & Minimum Feature Size: Select the operational laser band (e.g., 532 nm vs. 1550 nm) to define the etch depth ceiling h_max = {lambda} / {n - 1} and establish minimum feature pitch limits.
Execute Error-Tolerance Sweep: Run a Monte Carlo loop on the ASM code with +/- 5 mu m lateral plate jitter and +/- 5% etch depth over-etching to map system degradation envelopes.
Export Height Profiles to CAD/Mask Files: Convert the final 2D phase arrays phi_k(x, y) into 16-bit TIFF depth maps or GDSII layer files for lithographic mask generation.
From <https://gemini.google.com/app/cf38837645976e2f> Google Gemini (3.8 Flash)
Comments
Post a Comment