AI Green Solvent Substitution for Chemistry Projects

ISEF Category: Chemistry

Subcategory: Other  ·  Difficulty: Advanced  ·  Setup: School Lab  ·  Time: 1 to 2 Months

The Hook

A lot of chemistry still runs on solvents that are unsafe, expensive, or hard to dispose of. You can build a project that asks a smart question, can AI help you predict better replacements? Then you can test those ideas against a real solvent guide and a small experiment set. That gives you chemistry, data, and a practical problem in one project.

What Is It?

A solvent is the liquid that helps other substances dissolve, mix, or react. Think of it like the stage where the chemistry happens. Some solvents are useful, but they can also be flammable, toxic, or bad for the environment. Green chemistry tries to replace them with safer options.

This project pairs two things. First, you ask an AI model to suggest possible solvent substitutes for a reaction, extraction, or cleaning step. Second, you check those ideas against a trusted solvent-selection guide, such as the GSK guide, which ranks solvents by safety and environmental factors. Then you compare the AI's suggestions with a small experimental panel or literature-based test set. The goal is not to trust AI blindly. The goal is to see when it helps, when it misses, and which rules make its suggestions better.

Why This Is a Good Topic

This is a strong science fair topic because you can test it with clear criteria. You are not just asking whether AI sounds smart. You are checking whether it can make useful chemistry predictions and whether those predictions match a real safety-and-performance framework. That gives you measurable outputs, like agreement scores, ranking accuracy, and success rates on a small sample set. You also learn how chemists think about tradeoffs, not just yes or no answers.

Research Questions

• How does the wording of the prompt change the quality of AI solvent-substitution suggestions?
• What is the effect of adding the GSK solvent-selection guide to the prompt on the agreement between AI output and expert rankings?
• Does AI predict safer solvent replacements more accurately for polar solvents than for nonpolar solvents?
• To what extent do AI suggestions match the GSK guide when the target task is extraction versus reaction cleanup?
• Which prompt format produces the highest overlap between AI-ranked solvent options and guide-ranked solvent options?
• How does a small experimental panel change the final solvent ranking compared with AI-only ranking?

Basic Materials

• Laptop or desktop computer with internet access.
• Access to an AI chat model that can generate text responses.
• Printed or digital copy of the GSK solvent-selection guide.
• Spreadsheet software such as Google Sheets or Microsoft Excel.
• Notebook for logging prompts, outputs, and scores.
• Basic chemistry reference sources for checking solvent properties.
• A small set of safe, school-approved solvents or solvent analogs for comparison.
• PPE for supervised school lab work, including goggles, gloves, and a lab coat.

Advanced Materials

• University or school lab access with approved solvent handling.
• Analytical balance for mass measurements.
• UV-Vis spectrophotometer or other relevant assay equipment.
• GC-MS or HPLC access for composition or purity checks.
• Fume hood for solvent work.
• Reference standards for the target system.
• Controlled set of candidate green solvents and conventional solvents.
• Data logging system for repeated measurements and statistical comparison.

Software & Tools

• Google Sheets: Organizes AI outputs, guide scores, and experimental results in one table.
• Python: Helps you clean data, score agreement, and make comparison plots.
• Jupyter Notebook: Keeps your code, notes, and graphs in one shareable file.
• ImageJ: Measures visible color or intensity changes if your assay produces images.
• ChatGPT or another large language model: Generates candidate solvent substitutions and written rationales for comparison.

Experiment Steps

1. Define one solvent-use case, such as a reaction, extraction, or cleanup step, and keep the target chemistry narrow.
2. Build a scoring plan that turns solvent choices into numbers using safety, performance, and accessibility criteria.
3. Design prompt versions that differ in how much guidance they give the AI model.
4. Cross-check each AI suggestion against the GSK solvent-selection guide and record where they agree or conflict.
5. Plan a small validation set, using a few school-safe systems or literature-backed examples, to test whether the suggestion works in practice.
6. Decide in advance how you will compare AI-only ranking, guide-only ranking, and your final combined ranking.

Common Pitfalls

• Using a broad solvent category like 'better solvent' instead of a specific task, which makes the AI output too vague to judge.
• Comparing solvents without a clear scoring rubric, which turns the project into opinions instead of data.
• Treating the AI answer as a fact source, which can hide wrong chemistry or unsafe suggestions.
• Mixing different chemistry tasks in one dataset, which makes the guide comparison inconsistent.
• Testing too many solvents at once, which makes the project hard to analyze and weakens the conclusion.

What Makes This Competitive

A competitive version of this project needs more than a simple chatbot demo. You need a clear scoring system, a real comparison against a trusted solvent framework, and a test set that shows when AI succeeds or fails. Strong projects also separate chemistry accuracy from writing quality, since a model can sound confident and still be wrong. If you add statistical comparison and a thoughtful error analysis, your work starts to look like real research.

Project Variations

• Test whether AI does better with one-step prompt instructions or with a full solvent-ranking rubric.
• Compare AI solvent suggestions for organic synthesis, extraction, and purification tasks.
• Analyze whether the model favors commonly used solvents over greener but less familiar options.

Learn More

• GSK solvent-selection guide: Search for the official GSK solvent-selection guide PDF and use it to rank solvents by safety and environmental impact.
• ACS Green Chemistry Institute: Find articles, tools, and solvent-related resources on the American Chemical Society green chemistry pages.
• NIH PubMed: Search for review articles on green solvents, solvent replacement, and solvent selection metrics.
• Green Chemistry: Search the journal for solvent selection, solvent replacement, and machine-learning papers.
• MIT OpenCourseWare: Look for chemistry and data analysis course materials that help with experimental design and statistical thinking.