Multi-Step Workflows Tutorial

Learn how to combine multiple OHMind agents for complex, end-to-end computational chemistry workflows.

Table of Contents

Overview

Difficulty: 🔴 Advanced
Time: 90 minutes
Requirements: All external software configured (ORCA, GROMACS, Multiwfn)

In this tutorial, you will:

  1. Chain multiple agents for complex tasks
  2. Build end-to-end design-to-validation pipelines
  3. Combine computational methods for comprehensive analysis
  4. Learn workflow orchestration best practices

Why Multi-Step Workflows?

Real research problems often require multiple computational approaches:

graph LR
    A[Literature] --> B[Design]
    B --> C[QM Validation]
    C --> D[MD Simulation]
    D --> E[Analysis]
    E --> F[Report]

OHMind’s multi-agent architecture enables seamless transitions between:

  • RAG Agent → Literature search and synthesis
  • HEM Agent → Molecular design and optimization
  • Chemistry Agent → Structure manipulation
  • QM Agent → Electronic structure calculations
  • MD Agent → Dynamics simulations
  • Multiwfn Agent → Wavefunction analysis

Prerequisites

Full System Configuration

Ensure all components are configured:

# Check all paths
echo "ORCA: $OHMind_ORCA"
echo "Multiwfn: $MULTIWFN_PATH"
echo "Workspace: $OHMind_workspace"

# Verify GROMACS
gmx --version

# Check Qdrant (for RAG)
echo "Qdrant: ${QDRANT_PATH:-$QDRANT_URL}"

Start the Interface 

cd OHMind
./start_OHMind_cli.sh

Workflow 1: Literature-Guided Design

Objective

Use literature insights to guide HEM optimization, then validate top candidates.

Step 1: Literature Research

Prompt:

Search the literature for the most stable cation types for AEMs.
Identify:
1. Top 3 cation families by stability
2. Key structural features for stability
3. Recommended design principles

Expected Output:

Literature Analysis:

Top Cation Families:
1. Spirocyclic piperidinium (best stability)
2. N-alkyl piperidinium (good balance)
3. Sterically hindered imidazolium (good conductivity)

Key Structural Features:
- No β-hydrogens on N-substituents
- Steric protection around nitrogen
- Saturated ring systems preferred

Design Principles:
- Use bulky alkyl groups (ethyl, isopropyl)
- Avoid benzylic positions
- Consider spirocyclic architectures

Step 2: Guided Optimization

Prompt:

Based on the literature findings, run HEM optimization for:
- Backbone: PBF_BB_1
- Cation type: piperidinium (literature-recommended)
- Property: multi-objective
- Particles: 200
- Steps: 10

Focus on candidates with structural features identified in literature.

Expected Duration: 15-20 minutes

Step 3: QM Validation

Prompt:

For the top 3 candidates from the optimization:
1. Run geometry optimization at B3LYP/def2-SVP
2. Calculate LUMO energies
3. Compare with literature stability criteria (LUMO > -2.0 eV)
4. Rank by predicted stability

Expected Output:

Rank SMILES LUMO (eV) Stability
1 CC[N+]1(CC)CCCCC1 -1.85 ✓ Good
2 C[N+]1(CC)CCCCC1 -2.05 ~ Marginal
3 C[N+]1(C)CCCCC1 -2.25 ✗ Concerning

Step 4: Final Report

Prompt:

Summarize the literature-guided design workflow:
1. Literature insights used
2. Optimization results
3. QM validation findings
4. Final recommendations

Format as a brief research report.

Workflow 2: Design-to-Validation Pipeline

Objective

Complete pipeline from molecular design through multi-level validation.

Step 1: Initial Design

Prompt:

Design a new piperidinium cation for AEMs with these requirements:
1. High predicted alkaline stability
2. Molecular weight < 200 g/mol
3. Synthesizable from common precursors

Generate 5 candidate structures with SMILES.

Step 2: Property Screening

Prompt:

For the 5 candidate cations:
1. Calculate molecular weight
2. Estimate synthetic accessibility
3. Check for problematic functional groups
4. Rank by overall suitability

Eliminate any that don't meet criteria.

Step 3: QM Characterization

Prompt:

For the remaining candidates, run QM calculations:
1. Geometry optimization (B3LYP/def2-SVP)
2. Frequency calculation (verify minimum)
3. Property extraction:
   - HOMO/LUMO energies
   - Charges on nitrogen
   - Dipole moment

Create a comparison table.

Expected Duration: 30-45 minutes for multiple molecules

Step 4: Detailed Analysis

Prompt:

For the best candidate from QM screening:
1. Run Multiwfn analysis on the wavefunction
2. Visualize LUMO orbital
3. Identify potential degradation sites
4. Calculate electrophilicity index

Step 5: MD Validation

Prompt:

Build a small AEM system with the best candidate:
1. Create polymer (DP=10, 3 chains)
2. Add water (λ=15)
3. Run 500 ps MD at 400 K
4. Calculate OH⁻ diffusion coefficient
5. Estimate conductivity

Expected Duration: 30-45 minutes

Step 6: Comprehensive Report

Prompt:

Generate a comprehensive validation report for the best candidate:

Include:
1. Structure and properties
2. QM results (energies, orbitals, charges)
3. Wavefunction analysis (degradation sites)
4. MD results (diffusion, conductivity)
5. Comparison with literature values
6. Recommendation for experimental synthesis

Workflow 3: Comprehensive Property Prediction

Objective

Multi-scale property prediction combining QM and MD methods.

Step 1: Select Target

Prompt:

I want to comprehensively characterize this cation for AEM use:
SMILES: CC[N+]1(CC)CCCCC1 (N,N-diethylpiperidinium)

Create a characterization plan covering:
1. Electronic properties (QM)
2. Thermodynamic properties (QM)
3. Transport properties (MD)
4. Stability assessment

Step 2: Electronic Structure

Prompt:

Run electronic structure calculations:
1. Optimize geometry at B3LYP-D3BJ/def2-TZVP
2. Calculate:
   - Frontier orbital energies
   - Atomic charges (Hirshfeld)
   - Dipole and quadrupole moments
   - Electrostatic potential

Step 3: Thermochemistry

Prompt:

Calculate thermochemical properties:
1. Frequency calculation for ZPE
2. Gibbs free energy at 298 K and 400 K
3. Proton affinity (if applicable)
4. Solvation energy in water

Step 4: Wavefunction Analysis

Prompt:

Perform detailed wavefunction analysis with Multiwfn:
1. Natural population analysis
2. Bond order analysis
3. Electron localization function (ELF)
4. Identify reactive sites

Step 5: Transport Properties

Prompt:

Run MD simulations for transport properties:
1. Build hydrated system (λ=20)
2. Equilibrate at 400 K
3. Production run (1 ns)
4. Calculate:
   - OH⁻ diffusion coefficient
   - Water diffusion coefficient
   - Ionic conductivity (Nernst-Einstein)

Step 6: Stability Assessment

Prompt:

Assess alkaline stability based on all calculations:
1. LUMO energy and localization (QM)
2. Charge distribution (QM/Multiwfn)
3. Reactive site identification (Multiwfn)
4. Compare with stable reference compounds

Provide stability rating and confidence level.

Step 7: Property Summary

Prompt:

Create a comprehensive property card for the cation:

| Property | Value | Method | Confidence |
|----------|-------|--------|------------|
| LUMO | ? eV | DFT | High |
| Conductivity | ? mS/cm | MD | Medium |
| Stability | ? | Multi-method | High |
...

Include comparison with literature benchmarks.

Workflow 4: Comparative Analysis

Objective

Compare multiple cation designs using consistent methodology.

Step 1: Define Candidates

Prompt:

I want to compare these 4 cation types for AEM applications:
1. Piperidinium: C[N+]1(C)CCCCC1
2. Pyrrolidinium: C[N+]1(C)CCCC1
3. Imidazolium: Cn1cc[n+](C)c1
4. Morpholinium: C[N+]1(C)CCOCC1

Create a systematic comparison plan.

Step 2: Consistent QM Calculations

Prompt:

For all 4 cations, run identical QM calculations:
- Method: B3LYP-D3BJ/def2-SVP
- Geometry optimization
- Frequency verification
- Property extraction

Ensure consistent settings for fair comparison.

Step 3: Parallel MD Simulations

Prompt:

For all 4 cations, run comparable MD simulations:
- Same polymer backbone (PBF_BB_1)
- Same DP (15) and chain count (3)
- Same water content (λ=15)
- Same temperature (400 K)
- Same duration (500 ps)

Extract diffusion coefficients for comparison.

Step 4: Comparative Analysis

Prompt:

Create a comprehensive comparison of all 4 cations:

| Property | Piperidinium | Pyrrolidinium | Imidazolium | Morpholinium |
|----------|--------------|---------------|-------------|--------------|
| LUMO (eV) | | | | |
| N charge | | | | |
| D(OH⁻) | | | | |
| σ (mS/cm) | | | | |
| Stability | | | | |

Identify the best candidate for each property and overall.

Step 5: Trade-off Analysis

Prompt:

Analyze the trade-offs between the cation types:
1. Stability vs. conductivity
2. Synthesis complexity vs. performance
3. Cost vs. benefit

Create a recommendation matrix for different applications:
- High-stability applications
- High-conductivity applications
- Cost-sensitive applications

Best Practices

Workflow Design

  1. Plan ahead - Outline all steps before starting
  2. Save checkpoints - Note intermediate results
  3. Use consistent methods - Same level of theory for comparisons
  4. Document everything - Keep prompts and outputs

Prompt Engineering

Good multi-step prompt:

I'm running a design-to-validation workflow. We've completed:
1. ✓ Literature search (piperidinium recommended)
2. ✓ HEM optimization (top candidate: CC[N+]1(CC)CCCCC1)
3. Current: QM validation

For step 3, run geometry optimization and calculate LUMO energy.
Then we'll proceed to step 4 (MD simulation).

Avoid:

Do everything at once - optimize, run QM, run MD, analyze.

Resource Management

Task Typical Time Memory
HEM optimization 10-20 min 4 GB
QM optimization 5-15 min 2 GB
QM frequencies 10-30 min 4 GB
MD equilibration 10-20 min 2 GB
MD production 30-60 min 2 GB

Error Recovery

If a step fails:

Step 3 (QM calculation) failed with SCF convergence error.
Let's try:
1. Use a different initial guess
2. Increase SCF iterations
3. Try a smaller basis set first

Then continue with the workflow.

Troubleshooting

Common Multi-Step Issues

Issue Cause Solution
Lost context Long workflow Summarize progress periodically
Inconsistent results Different methods Use same settings throughout
Timeout Long calculations Break into smaller steps
Missing files Workflow interruption Check workspace for intermediates

Workflow Recovery 

My workflow was interrupted after step 3. The QM results are in 
$QM_WORK_DIR/temp_abc123/. Resume from step 4 using those results.

Debugging Complex Workflows 

The MD simulation in my workflow gave unexpected results.
Show me:
1. The system composition
2. The simulation parameters
3. The energy profile

Let's diagnose the issue before continuing.

See Also

*Last updated: 2025-12-23 OHMind v1.0.0*

PolyAI Team
Copyright © 2009-2025 Changchun Institute of Applied Chemistry, Chinese Academy of Sciences
Address: No. 5625, Renmin Street, Changchun, Jilin, China. Postal Code: 130022