HEM Optimization Tutorial

Learn how to design new cations for Hydroxide Exchange Membranes using Particle Swarm Optimization (PSO).

Overview
Prerequisites
Part 1: Exploring the Design Space
Part 2: Running an Optimization
Part 3: Analyzing Results
Part 4: Advanced Optimization
Expected Outputs
Troubleshooting
Next Steps
See Also

Overview

Difficulty: 🟢 Beginner
Time: 30 minutes
Requirements: OHMind installed, LLM configured

In this tutorial, you will:

Explore available backbones and cation types
Run a PSO optimization to design new cations
Analyze and interpret the optimization results
Learn advanced optimization techniques

What is HEM Optimization?

Hydroxide Exchange Membranes (HEMs) are critical components in fuel cells and electrolyzers. The performance of HEMs depends heavily on the cation structure attached to the polymer backbone. OHMind uses Particle Swarm Optimization (PSO) combined with a Junction Tree VAE to explore the chemical space and find optimal cation structures.

Optimization Objectives

OHMind can optimize for multiple properties:

Property	Code	Description
Electrical Conductivity	`ec`	Ion transport efficiency
Electrochemical Stability Range	`esr`	Voltage window stability
Excess Water Uptake	`ewu`	Water management
Multi-objective	`multi`	Balanced optimization of all properties

Prerequisites

System Setup

# Activate environment
conda activate OHMind

# Verify workspace
echo $OHMind_workspace
# Should show your workspace path

# Start the CLI
./start_OHMind_cli.sh

Verify HEM Agent

The HEM agent should be available. You can verify by asking:

What agents and tools do you have available for HEM design?

Expected response: The agent should list HEMDesign tools including ListBackbones, ListCations, HEMOptimizer, etc.

Part 1: Exploring the Design Space

Step 1.1: List Available Backbones

First, let’s see what polymer backbones are available for HEM design.

Prompt:

Using your HEMDesign tools, list all available polymer backbones 
for HEM design. Show their names and SMILES structures.

Expected Output:

The agent will return a table of backbones:

Backbone	SMILES	Description
PBF_BB_1	`[]c1ccc(-c2ccc([])cc2)cc1`	Polybiphenyl fluorene
PP_BB_1	`[]c1ccc([])cc1`	Polyphenylene
PPO_BB_1	`[]Oc1ccc([])cc1`	Poly(phenylene oxide)
…	…	…

Step 1.2: List Available Cation Types

Next, explore the cation families that can be optimized.

Prompt:

List all available cation types for HEM optimization. 
Show their names and initial SMILES templates.

Expected Output:

Cation Type	Initial SMILES	Description
piperidinium	`C[N+]1(C)CCCCC1`	6-membered ring
imidazolium	`Cn1cc[n+](C)c1`	Aromatic 5-membered
pyrrolidinium	`C[N+]1(C)CCCC1`	5-membered ring
…	…	…

Step 1.3: Validate a Configuration

Before running optimization, validate your chosen configuration.

Prompt:

Validate an optimization configuration for:
- Backbone: PBF_BB_1
- Cation: piperidinium
- Property: multi (multi-objective)

Is this a valid combination?

Expected Output:

The agent will confirm whether the configuration is valid and explain any constraints.

Part 2: Running an Optimization

Step 2.1: Basic Optimization

Now let’s run an actual optimization. Start with a small run to verify everything works.

Prompt:

Run a HEM optimization with these settings:
- Backbone: PBF_BB_1
- Cation type: piperidinium
- Property: multi (multi-objective)
- Particles: 100
- Steps: 3

This is a test run to verify the system works.

What Happens:

The HEM agent receives your request
It calls the HEMOptimizer tool
PSO optimization runs in the background
Progress is logged to the workspace
Results are saved as CSV files

Expected Duration: 2-5 minutes for a small run

Step 2.2: Monitor Progress

While optimization runs, you can check progress.

Prompt:

Show me the latest log entries for the current HEM optimization.

Expected Output:

Step 1/3, max: 0.45, min: 0.12, mean: 0.28
Step 2/3, max: 0.52, min: 0.15, mean: 0.31
Step 3/3, max: 0.58, min: 0.18, mean: 0.35

Step 2.3: Full Optimization Run

Once verified, run a more thorough optimization.

Prompt:

Run a full HEM optimization with these settings:
- Backbone: PBF_BB_1
- Cation type: piperidinium
- Property: multi (multi-objective)
- Particles: 200
- Steps: 10

Save results to the default workspace location.

Expected Duration: 10-20 minutes

Part 3: Analyzing Results

Step 3.1: Check Optimization Status

After the run completes, check the results.

Prompt:

Check the status of HEM optimization for PBF_BB_1 with piperidinium.
Show me the top 10 candidate molecules with their fitness scores.

Expected Output:

Rank	SMILES	EC Score	EWU Score	ESR Score	Overall
1	`CC[N+]1(CC)CCCCC1`	0.82	0.75	0.88	0.81
2	`C[N+]1(CC)CCCCC1`	0.79	0.78	0.85	0.80
3	…	…	…	…	…

Step 3.2: Interpret Results

Ask the agent to explain the results.

Prompt:

Analyze the top 5 candidates from the PBF_BB_1 piperidinium optimization.
Explain:
1. What structural features make them good candidates?
2. How do their predicted properties compare?
3. Which would you recommend for experimental validation?

Expected Output:

The agent will provide:

Structural analysis of top candidates
Property comparisons
Recommendations with reasoning

Step 3.3: Visualize Candidates

Generate structure images for top candidates.

Prompt:

For the top 3 candidates from the optimization, generate 2D structure 
images and save them to the workspace.

Expected Output:

Images saved to $OHMind_workspace/HEM/ or displayed in the interface.

Part 4: Advanced Optimization

Step 4.1: Single-Objective Optimization

Focus on a specific property.

Prompt:

Run a HEM optimization focused only on electrical conductivity (EC):
- Backbone: PP_BB_1
- Cation type: imidazolium
- Property: ec
- Particles: 150
- Steps: 8

Step 4.2: Comparative Study

Compare different backbone-cation combinations.

Prompt:

Compare the optimization results for these combinations:
1. PBF_BB_1 + piperidinium
2. PP_BB_1 + piperidinium
3. PBF_BB_1 + imidazolium

For each, run a quick optimization (100 particles, 5 steps) and 
summarize which combination shows the most promise.

Step 4.3: Constrained Optimization

Add constraints to the optimization.

Prompt:

Run a HEM optimization for PBF_BB_1 with piperidinium, but focus on 
candidates that:
1. Have molecular weight under 200 g/mol
2. Show high alkaline stability (ESR > 0.7)
3. Maintain reasonable conductivity (EC > 0.5)

Use multi-objective optimization with 200 particles for 10 steps.

Expected Outputs

File Locations

After optimization, find results in:

$OHMind_workspace/HEM/
├── best_solutions_PBF_BB_1_piperidinium.csv
├── best_fitness_history_PBF_BB_1_piperidinium.csv
└── optimization_PBF_BB_1_piperidinium.log

CSV File Format

best_solutions_*.csv:

Column	Description
smiles	Candidate SMILES
ec_score	Electrical conductivity score
ewu_score	Water uptake score
esr_score	Stability score
overall_fitness	Combined fitness
step	Optimization step found

best_fitness_history_*.csv:

Column	Description
step	Optimization step
max_fitness	Best fitness in swarm
min_fitness	Worst fitness in swarm
mean_fitness	Average fitness

Interpreting Scores

Score Range	Interpretation
0.8 - 1.0	Excellent candidate
0.6 - 0.8	Good candidate
0.4 - 0.6	Moderate candidate
< 0.4	Poor candidate

Troubleshooting

Common Issues

Issue	Cause	Solution
“Backbone not found”	Invalid backbone name	Use `ListBackbones` to see valid names
“Optimization timeout”	Long computation	Increase timeout in mcp.json
“No results found”	Optimization didn’t complete	Check logs with `ShowLogs`
“Permission denied”	Workspace not writable	Check `$HEM_SAVE_PATH` permissions

Checking Logs

Show me the full optimization log for PBF_BB_1 piperidinium.

Killing Stuck Jobs

List any running HEM optimization jobs and kill any that appear stuck.

Next Steps

After completing this tutorial:

Try different combinations - Explore other backbone-cation pairs
Validate with QM - Use QM Calculations to verify top candidates
Run MD simulations - Use MD Simulations for property validation
Search literature - Use Literature Search to compare with known compounds

Suggested Follow-up Prompts

For the top candidate from my optimization, run a QM calculation 
to verify its LUMO energy and alkaline stability.

Search the literature for similar cation structures and compare 
their reported properties with my optimization predictions.

HEM Optimization Tutorial

Table of Contents

Overview

What is HEM Optimization?

Optimization Objectives

Prerequisites

System Setup

Verify HEM Agent

Part 1: Exploring the Design Space

Step 1.1: List Available Backbones

Step 1.2: List Available Cation Types

Step 1.3: Validate a Configuration

Part 2: Running an Optimization

Step 2.1: Basic Optimization

Step 2.2: Monitor Progress

Step 2.3: Full Optimization Run

Part 3: Analyzing Results

Step 3.1: Check Optimization Status

Step 3.2: Interpret Results

Step 3.3: Visualize Candidates

Part 4: Advanced Optimization

Step 4.1: Single-Objective Optimization

Step 4.2: Comparative Study

Step 4.3: Constrained Optimization

Expected Outputs

File Locations

CSV File Format

Interpreting Scores

Troubleshooting

Common Issues

Checking Logs

Killing Stuck Jobs

Next Steps

Suggested Follow-up Prompts

See Also