{"id":216988,"date":"2022-06-28T04:53:51","date_gmt":"2022-06-28T04:53:51","guid":{"rendered":"\/forum\/forums\/topic\/inverse-design-of-polymer-y-branch-modify-the-example-provided\/"},"modified":"2023-04-18T06:46:08","modified_gmt":"2023-04-18T06:46:08","slug":"inverse-design-of-polymer-y-branch-modify-the-example-provided","status":"closed","type":"topic","link":"https:\/\/innovationspace.ansys.com\/forum\/forums\/topic\/inverse-design-of-polymer-y-branch-modify-the-example-provided\/","title":{"rendered":"Inverse design of polymer y-branch – modify the example provided"},"content":{"rendered":"

Hi all,<\/p>\n

I tried to optimize a polymer y-branch according to the example here (https:\/\/optics.ansys.com\/hc\/en-us\/articles\/360042305274).<\/p>\n

Since polymer (SU-8) has a much lower refractive index, the main different is the size. I modified all sizes in the base setup .py file and the simulation .py file, yet error occured saying “The initial X is infeasible.  Restart with its projection.” As a result, the polygon structure extends beyond the bounds defined, and FoM could not be calculated.<\/p>\n

Attached please find the screenshot and codes. Please help to check if I made anything wrong.<\/p>\n

Thank you very much!<\/p>\n

[File 1]<\/p>\n

#############################################################################<\/p>\n

# Scriptfile: y_branch_opt_2D.py<\/p>\n

#<\/p>\n

# Description:<\/p>\n

# This script sets up and runs the adjoint shape-based optimization for inverse<\/p>\n

# design of the SOI Y-branch in 2D<\/p>\n

#<\/p>\n

# Steps include:<\/p>\n

# 1. Define the base simulation<\/p>\n

# 2. Define the optimizable geometry and optimization parameters<\/p>\n

# 3. Run optimization<\/p>\n

# 4. Save results<\/p>\n

#<\/p>\n

# Copyright 2019, Lumerical Solutions, Inc.<\/p>\n

# Copyright chriskeraly<\/p>\n

##############################################################################<\/p>\n

 <\/p>\n

import os, sys<\/p>\n

import numpy as np<\/p>\n

import scipy as sp<\/p>\n

import lumapi<\/p>\n

 <\/p>\n

from lumopt.utilities.wavelengths import Wavelengths<\/p>\n

from lumopt.geometries.polygon import FunctionDefinedPolygon<\/p>\n

from lumopt.utilities.materials import Material<\/p>\n

from lumopt.figures_of_merit.modematch import ModeMatch<\/p>\n

from lumopt.optimizers.generic_optimizers import ScipyOptimizers<\/p>\n

from lumopt.optimization import Optimization<\/p>\n

 <\/p>\n

######## BASE SIMULATION ########<\/p>\n

sys.path.append(os.path.dirname(__file__))<\/p>\n

from varFDTD_y_branch import y_branch_init_<\/p>\n

y_branch_base = y_branch_init_<\/p>\n

 <\/p>\n

######## DIRECTORY FOR GDS EXPORT #########<\/p>\n

example_directory = os.getcwd()<\/p>\n

 <\/p>\n

######## SPECTRAL RANGE #########<\/p>\n

wavelengths = Wavelengths(start = 1530e-9, stop = 1570e-9, points = 21)<\/p>\n

 <\/p>\n

######## DEFINE OPTIMIZABLE GEOMETRY ########<\/p>\n

# The class FunctionDefinedPolygon needs a parameterized Polygon (with points ordered<\/p>\n

# in a counter-clockwise direction). Here the geometry is defined by 10 parameters defining<\/p>\n

# the knots of a spline, and the resulting Polygon has 200 edges, making it quite smooth.<\/p>\n

 <\/p>\n

# Define the span and number of points<\/p>\n

initial_points_x = np.linspace(-12.5e-6, 12.5e-6, 10)<\/p>\n

initial_points_y = np.linspace(5e-6, 5e-6, initial_points_x.size)<\/p>\n

def splitter(params):<\/p>\n

”’ Defines a taper where the paramaters are the y coordinates of the nodes of a cubic spline. ”’<\/p>\n

 <\/p>\n

## Include two set points based on the initial guess. The should attach the optimizeable geometry to the input and output<\/p>\n

points_x = np.concatenate(([initial_points_x.min() – 0.01e-6], initial_points_x, [initial_points_x.max() + 0.01e-6]))<\/p>\n

points_y = np.concatenate(([initial_points_y.min()], params, [initial_points_y.max()]))<\/p>\n

 <\/p>\n

## Up sample the polygon points for a smoother curve. Some care should be taken with interp1d object. Higher degree fit<\/p>\n

# “cubic”, and “quadratic” can vary outside of the footprint of the optimization. The parameters are bounded, but the<\/p>\n

# interpolation points are not. This can be particularly problematic around the set points.<\/p>\n

n_interpolation_points = 100<\/p>\n

polygon_points_x = np.linspace(min(points_x), max(points_x), n_interpolation_points)<\/p>\n

interpolator = sp.interpolate.interp1d(points_x, points_y, kind = ‘cubic’)<\/p>\n

polygon_points_y = interpolator(polygon_points_x)<\/p>\n

print(polygon_points_y)<\/p>\n

 <\/p>\n

### Zip coordinates into a list of tuples, reflect and reorder. Need to be passed ordered in a CCW sense<\/p>\n

polygon_points_up = [(x, y) for x, y in zip(polygon_points_x, polygon_points_y)]<\/p>\n

polygon_points_down = [(x, -y) for x, y in zip(polygon_points_x, polygon_points_y)]<\/p>\n

polygon_points = np.array(polygon_points_up[::-1] + polygon_points_down)<\/p>\n

return polygon_points<\/p>\n

 <\/p>\n

# L-BFGS methods allows the parameters to be bound. These should enforse the optimization footprint defined in the setup<\/p>\n

bounds = [(0.2e-6, 0.8e-6)] * initial_points_y.size<\/p>\n

print(“Bounds:”)<\/p>\n

print(bounds)<\/p>\n

 <\/p>\n

#Load from 2D results if availble<\/p>\n

try:<\/p>\n

prev_results = np.loadtxt(‘2D_parameters.txt’)<\/p>\n

except:<\/p>\n

print(“Couldn’t find the file containing 2D optimization parameters. Starting with default parameters”)<\/p>\n

prev_results = initial_points_y<\/p>\n

 <\/p>\n

# Set device and cladding materials, as well as as device layer thickness<\/p>\n

eps_in = Material(name = ‘SU-8’)<\/p>\n

eps_out = Material(name = ‘air’)<\/p>\n

depth = 2e-6<\/p>\n

 <\/p>\n

# Initialize FunctionDefinedPolygon class<\/p>\n

polygon = FunctionDefinedPolygon(func = splitter,<\/p>\n

initial_params = prev_results,<\/p>\n

bounds = bounds,<\/p>\n

z = 0.0,<\/p>\n

depth = depth,<\/p>\n

eps_out = eps_out, eps_in = eps_in,<\/p>\n

edge_precision = 5,<\/p>\n

dx = 1.0e-9)<\/p>\n

 <\/p>\n

 <\/p>\n

######## FIGURE OF MERIT ########<\/p>\n

fom = ModeMatch(monitor_name = ‘fom’,<\/p>\n

mode_number = ‘fundamental mode’,<\/p>\n

direction = ‘Forward’,<\/p>\n

target_T_fwd = lambda wl: np.ones(wl.size),<\/p>\n

norm_p = 1)<\/p>\n

 <\/p>\n

######## OPTIMIZATION ALGORITHM ########<\/p>\n

optimizer = ScipyOptimizers(max_iter = 30,<\/p>\n

method = ‘L-BFGS-B’,<\/p>\n

#scaling_factor = scaling_factor,<\/p>\n

pgtol = 1.0e-5,<\/p>\n

ftol = 1.0e-5,<\/p>\n

#target_fom = 0.0,<\/p>\n

scale_initial_gradient_to = 0.0)<\/p>\n

 <\/p>\n

######## PUT EVERYTHING TOGETHER ########<\/p>\n

opt = Optimization(base_script = y_branch_base,<\/p>\n

wavelengths = wavelengths,<\/p>\n

fom = fom,<\/p>\n

geometry = polygon,<\/p>\n

optimizer = optimizer,<\/p>\n

use_var_fdtd = True,<\/p>\n

hide_fdtd_cad = False,<\/p>\n

use_deps = True,<\/p>\n

plot_history = True,<\/p>\n

store_all_simulations = False)<\/p>\n

 <\/p>\n

######## RUN THE OPTIMIZATION ########<\/p>\n

results = opt.run()<\/p>\n

 <\/p>\n

######## SAVE THE BEST PARAMETERS TO FILE ########<\/p>\n

np.savetxt(‘..\/2D_parameters.txt’, results[1])<\/p>\n

 <\/p>\n

######## EXPORT OPTIMIZED STRUCTURE TO GDS ########<\/p>\n

gds_export_script = str(“gds_filename = ‘y_branch_2D.gds’;” +<\/p>\n

“top_cell = ‘model’;” +<\/p>\n

“layer_def = [1, {0}, {1}];”.format(-depth\/2, depth\/2) +<\/p>\n

“n_circle = 64;” +<\/p>\n

“n_ring = 64;” +<\/p>\n

“n_custom = 64;” +<\/p>\n

“n_wg = 64;” +<\/p>\n

“round_to_nm = 1;” +<\/p>\n

“grid = 1e-9;” +<\/p>\n

“max_objects = 10000;” +<\/p>\n

“Lumerical_GDS_auto_export;”)<\/p>\n

 <\/p>\n

with lumapi.MODE(hide = False) as mode:<\/p>\n

mode.cd(example_directory)<\/p>\n

y_branch_init_(mode)<\/p>\n

mode.addpoly(vertices = splitter(results[1]))<\/p>\n

mode.set(‘x’, 0.0)<\/p>\n

mode.set(‘y’, 0.0)<\/p>\n

mode.set(‘z’, 0.0)<\/p>\n

mode.set(‘z span’, depth)<\/p>\n

mode.set(‘material’,’SU-8′)<\/p>\n

mode.save(“y_branch_2D_FINAL”)<\/p>\n

input(‘Enter…’)<\/p>\n

mode.eval(gds_export_script)<\/p>\n

 <\/p>\n

\"\"<\/a><\/p>\n

 <\/p>\n","protected":false},"template":"","class_list":["post-216988","topic","type-topic","status-closed","hentry","topic-tag-coupling","topic-tag-design-optimization","topic-tag-fdtd","topic-tag-optimization","topic-tag-polymer","topic-tag-waveguides"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t