Lumerical Forum

Ansys Lumerical Learning Forum is the primary hub for users to discuss simulation setups and troubleshoot discrepancies between their results and those found in "proper papers" (published research). Understanding "Proper Paper" Discrepancies

A common topic on the forum is why a user's simulation doesn't match a published paper. Experts typically point to several critical factors: Ansys Innovation Space Unique Software Features

: Lumerical has specific settings and algorithms that new users must learn before achieving reliable, paper-quality results. Published "Best" Results

: Research papers often present the most optimized results from hundreds of different simulation configurations. Sensitivity to Accuracy : Minor differences in mesh accuracy PML (Perfectly Matched Layer) thickness, or simulation time

can lead to significant deviations in sensitive measurements like phase shift or power. Ansys Innovation Space Key Forum Resources for Paper Reproduction Lumerical Learning Forum (Photonics)

: The main sub-forum for Q&A on FDTD, MODE, CHARGE, and other solvers. Ansys Insight: Result Discrepancies

: A dedicated article explaining why your simulation might differ from a peer-reviewed paper or experiment. Convergence Testing Guide lumerical forum

: Crucial documentation for ensuring your simulation has converged to a stable, accurate result suitable for publication. Ansys Innovation Space Specific Simulation Fixes

If you are struggling to match a paper's results, forum experts often recommend: Mesh Refinement

: Reducing "max edge length" or increasing "max refine steps" for complex structures. Material Modeling : Checking the multi-coefficient model fit

for dispersive materials to ensure the simulation accurately represents the real-world material. Source Configuration

: Ensuring correct usage of plane waves, dipoles, or integrated mode sources based on the paper's experimental setup. Ansys Optics Are you trying to reproduce a specific figure from a paper, and if so, what is the material or structure you are simulating?

Title: Optimizing FDTD Simulation of a Photonic Crystal Structure Ansys Lumerical Learning Forum is the primary hub

Description:

Hi everyone,

I'm working on simulating a photonic crystal structure using Lumerical's FDTD solver. I've been trying to optimize the simulation to achieve accurate results within a reasonable computation time. I'd love to hear from others who have experience with similar simulations.

My structure consists of a 2D array of air holes in a silicon background, with a lattice constant of 500 nm. I'm using a Gaussian source with a wavelength range of 1500-1600 nm. I've meshed the structure with a maximum mesh size of 20 nm.

To optimize the simulation, I've tried the following:

Using a coarse mesh size of 50 nm and then refining it to 20 nm in the regions of interest
Implementing symmetry to reduce the simulation size
Using a PML (perfectly matched layer) to minimize reflections

However, I'm still experiencing some issues with accuracy and computation time. Has anyone else encountered similar challenges? What strategies have you used to optimize your FDTD simulations? Using a coarse mesh size of 50 nm

Specifically, I'd like to know:

What are the best practices for meshing complex structures like photonic crystals?
How can I accurately model the material properties of my structure (e.g., silicon and air)?
Are there any specific optimization techniques that can be used to reduce computation time without sacrificing accuracy?

Any advice or insights would be greatly appreciated. I've attached my simulation file (.lms) for reference.

Attachments: photonic_crystal.lms

Tags: FDTD, photonic crystal, optimization

This post is well-written because it:

Clearly states the problem and the goal
Provides relevant background information (e.g., structure description, simulation settings)
Shows that effort has been made to solve the problem (e.g., meshing, symmetry, PML)
Specifically asks for advice or insights from others
Includes relevant attachments (e.g., simulation file) and tags

4. Searching Before Posting

80% of questions have already been answered.

Use keywords like: "PML reflection FDTD", "CHARGE convergence error", "optical generation rate".
Limit search to last 2 years for modern versions (2022 R1 onward).
Check pinned posts in each subforum – they contain common solutions and example files.

6. Using Attachments & Code Blocks

Scripts: Wrap in [code] ... [/code] or use triple backticks ```lua.
Images: Use Imgur or upload directly (prefer screenshots of material fitler, mesh, and log file).
Simulation files: Remove large output monitors before sharing.

2. Creating an Account

You need a free Ansys account to post or reply.
Sign up at: https://www.ansys.com/sign-in
After login, navigate to the Community → Forums section.

The Future of the Lumerical Forum Under Ansys

Since the acquisition, the forum has become more integrated with the broader Ansys ecosystem (including HFSS, Q3D Extractor, and Mechanical). Recent updates include a unified login with the Ansys Customer Portal and better support for multiphysics coupling. Expect to see deeper AI-assisted search within the forum—allowing you to ask natural language questions like, “How do I model a graphene-based modulator in FDTD?” and receive curated threads as answers.

How to Ask a Winning Question on the Lumerical Forum

Because the community is technical, the quality of the answer correlates directly with the quality of the question. To get a solution within hours (not days), follow these protocols:

Attach the .lms or .ldf file: If your simulation is proprietary, create a minimal working example (MWE). A moderator cannot debug a simulation they cannot run.
Specify the solver version: Lumerical updates frequently. Code that worked in version 2020 R2 may break in 2024 R1. Always state your build number.
Post the exact error log: Do not paraphrase. Copy the entire script output, including the time stamps and warning flags.
Show what you have already tried: This proves you aren’t asking others to do your homework. “I tried mesh refinement from 10 nm to 2 nm, but the Q-factor still saturates at 5,000” is a great opening.