Sources

Optical field sources for system initialization.

Overview

The Sources module provides components that generate optical fields to initialize optical systems. Sources can be static (fixed beam profile) or trainable (optimizable during inverse design).

Examples

Creating Sources

using FluxOptics

# Create spatial grid and mode
xv, yv = spatial_vectors(128, 128, 2.0, 2.0)
gaussian = Gaussian(20.0)
u0 = ScalarField(gaussian(xv, yv), (2.0, 2.0), 1.064)

# Static source (fixed beam profile)
source_static = ScalarSource(u0)

istrainable(source_static)  # false

false

# Trainable source (optimizable)
source_train = ScalarSource(u0; trainable=true, buffered=true)

(istrainable(source_train), isbuffered(source_train))

(true, true)

Using Sources in Systems

# Generate field from source
u_generated = propagate(source_static)

size(u_generated)  # Same as template

(128, 128)

# Build system with trainable source
phase_mask = Phase(u0, (x, y) -> 0.01*(x^2 + y^2))
propagator = ASProp(u0, 500.0)

system = source_train |> phase_mask |> propagator

# Execute system
result = system()
output = result.out

power(output)[]  # Propagated through system

1.0000000000000004

Accessing Source Data

# Extract current source field
current_field = get_source(source_train)

# Access trainable parameters (for optimization)
params = trainable(source_train)

keys(params)

(:u0,)

Key Types

ScalarSource: Source from scalar field template

Key Functions

get_source: Access source field data
propagate: Generate field from source

Usage Patterns

Static Sources

Use for fixed, known beam profiles:

source = ScalarSource(u0)  # trainable=false by default

Trainable Sources

Enable optimization of source amplitude and phase during inverse design:

source = ScalarSource(u0; trainable=true, buffered=true)
# ... optimization updates source field ...
optimized_beam = get_source(source)

Multi-Mode Sources

Sources support multi-mode fields:

# Create multi-mode field
modes = hermite_gaussian_groups(15.0, 3)
mode_data = stack([m(xv, yv) for m in modes], dims=3)
u_multi = ScalarField(mode_data, (2.0, 2.0), 1.064)

source_multi = ScalarSource(u_multi; trainable=true)