AI Scientist Platform: Phenylethyl Resorcinol Orbital & ESP

This AI Scientist Platform demonstration follows a phenylethyl resorcinol task through molecular-orbital and electrostatic-potential (ESP) analysis. Mira kept the project context, called the required Skills, ran the task, and retained conformers, DFT outputs, HOMO/LUMO data, and an ESP map. The calculation is a workflow demonstration, not a claim about a product, efficacy, or safety.

The useful part is the platform record around the phenylethyl resorcinol calculation. A chemist can inspect the input structure, method, task outputs, and limits of the conclusion instead of receiving an unsupported answer in a chat.

The workspace keeps the task prompt, project files, Skill calls, task status, and output artifacts together.

By Mira · Published: July 14, 2026

How this article was prepared: it is based on one completed Mira task and its output files. The calculation used a gas-phase model. Its results demonstrate a quantum chemistry workflow and do not establish a property of a named commercial ingredient, a finished product, efficacy, or safety.

---

The phenylethyl resorcinol calculation

This task asks Mira to examine molecular-orbital information and electrostatic potential for phenylethyl resorcinol. The run sampled 20 conformers, optimized the selected structure with wb97x-d4/def2-svpd, and produced HOMO/LUMO and ESP outputs. It is the concrete calculation that this article uses to show the platform workflow.

How the platform carried the calculation

A quantum chemistry question often crosses several tools and handoffs. A researcher needs to define the task, choose a starting geometry, run calculations, inspect the outputs, and decide whether another calculation or an experiment should follow.

Mira connected those steps through four platform capabilities:

• Project: the question, assumptions, and task history stayed in one research context.
• Skills: specialized capabilities prepared the molecular input, conformer search, calculation, and analysis steps.
• Task: one traceable execution carried the calculation from structure to results.
• Files: coordinates, logs, wavefunctions, CSV summaries, and figures remained with the task for review.

That record gives a chemistry team something to review and repeat. It does not transfer scientific responsibility to the platform.

The demonstration input and workflow

A resorcinol-derivative model compound was supplied as the molecular input: SMILES Oc1cc(O)cc(CCc2ccccc2)c1, molecular formula C₁₄H₁₄O₂, and a neutral singlet state. The input is shown so readers can see what the task calculated; it is not presented as the identity of a commercial ingredient. Those details matter because every later output depends on them.

The task then followed five steps:

1. Confirm the structure. Mira parsed the SMILES and produced a 2D structure image for inspection.
2. Sample conformers. CREST with GFN2-xTB generated 20 conformers. The lowest-energy conformer moved into the DFT stage.
3. Optimize and calculate. GPU4PySCF ran DFT geometry optimization and a single-point calculation at wb97x-d4/def2-svpd. The task exported a Molden wavefunction file for later analysis.
4. Inspect orbitals and electrostatic potential. Multiwfn generated HOMO/LUMO, ESP, and cube-file outputs. VMD rendered the orbital and ESP images.
5. Keep the run reviewable. Coordinates, logs, Molden wavefunctions, Multiwfn tables, cube files, VMD images, and rendering scripts remained attached to the task.

CREST, GPU4PySCF, Multiwfn, and VMD performed the scientific calculations and visualizations. Mira submitted work to compute nodes, handed intermediate files between steps, ran batch scripts, tracked the long task, and organized the resulting files. Researchers can inspect the calculation record or adjust visualization parameters themselves.

Results from the demonstration run

These values report what the task calculated for the supplied model compound. They do not describe product performance or biological activity.

The two hydroxyl oxygens appeared near electron-rich regions in this output, with average surface electrostatic potentials of about -18.6 and -18.5 kcal/mol. The calculated surface was approximately 58% positive electrostatic potential and 42% negative. A researcher can use those observations to choose a matched comparison molecule or plan a more specific interaction study.

A review checklist before using these numbers

The value of a result is not only the number on the page. Before using this run in a project discussion, a reviewer can check five things:

1. Identity: Does the name, SMILES, molecular formula, charge, and spin state match the compound being discussed?
2. Starting point: Was a conformer search performed, and is the selected geometry documented?
3. Method: Is the functional and basis set stated, with the calculation logs and wavefunction available?
4. Scope: Is the result described as a gas-phase electronic-structure observation, rather than an efficacy or safety result?
5. Comparison: Are control compounds calculated under the same assumptions before relative conclusions are drawn?

This checklist is deliberately simple. It prevents a visually persuasive orbital or surface map from becoming a conclusion that its model cannot support. It also gives another researcher a fast way to decide whether the task is reusable.

What the orbital and ESP outputs mean

HOMO shows where electron density may be comparatively available for an electron-donation discussion. LUMO provides a complementary view of electron-accepting character. Their energies and spatial distributions are most useful when a team calculates related structures at the same level of theory.

Molecular electrostatic potential (ESP) maps relative electrostatic environments on a molecular surface. The negative regions around the hydroxyl oxygens are a useful observation for this model compound. Solvent, pH, concentration, and other environmental conditions can change how a real system behaves, so the map does not establish a target interaction or a product benefit.

Turn one result into a matched comparison

A single molecule is a starting point. The more informative next step is a comparison that holds the model constant. A team can calculate related model compounds with the same structure checks, conformer strategy, charge and spin assumptions, and level of theory. It can then compare orbital distributions, gaps, and ESP patterns as model-specific observations.

That is not a ranking of compound quality. It is a way to make a research question sharper: which structural change is associated with the observed difference, and which experiment could test whether that difference matters in the intended system? Mira keeps the inputs, outputs, and conclusion alongside each other so the comparison does not become a spreadsheet of detached values.

When this protocol needs a different model

The right next calculation depends on the question, not on a fixed recipe. A gas-phase result can be suitable for an initial structure-level comparison. If the decision depends on solution behavior, ionization, a specific binding partner, or a formulation component, the team should state that requirement and choose a model that addresses it.

For example, a solvent model may be needed when comparing structures in a liquid environment. Alternative protonation states may need review if pH is relevant. A proposed interaction with another molecule calls for an explicit complex or another appropriate method, not an inference from an ESP image alone. Each change should create a new, traceable task with its own inputs, assumptions, and limitations. That makes the workflow more useful as the research question becomes more specific.

How a team can reuse the platform workflow

A task becomes more useful when its assumptions survive the handoff. A chemistry team can use this run as a starting protocol:

1. State the decision the calculation should inform.
2. Record the molecule, SMILES, charge, and spin state.
3. Preserve the conformer count, method, logs, wavefunction, and figures.
4. Run matched calculations for related model compounds.
5. Review whether a solvent model, a different method, or an experiment should follow.

PySCF and related electronic-structure tools make programmable calculations possible. Mira provides the project record around them: a place to keep inputs and outputs, call specialized Skills, run longer tasks, and give a reviewer the evidence with its limitations.

For a separate example of how model selection changes an agent workflow, read our GLM-5.2 free browser guide.

What remains with the researcher

The run cannot replace experimental validation, safety assessment, or a domain-specific model. A single gas-phase calculation also cannot confirm a target interaction, a real-world outcome, or regulatory suitability.

Researchers still decide whether the input is correct, whether the method fits the question, and whether the result supports the next step. Mira helps keep that decision trail intact. To run a chemistry question in the same project context, start a Mira project.

FAQ

What makes this an AI Scientist Platform demonstration rather than a chat answer?

The task retained the molecular input, computational steps, files, computed observations, and stated limits together. A researcher can inspect the run and use the same protocol for a comparison molecule.

Can this calculation prove a real-world property of the model compound?

No. It provides molecular-level information that can shape a testable question. Any real-world claim requires an appropriate model and experimental validation.

Why did the task sample conformers before DFT?

Flexible molecules can have several reasonable conformers. Sampling reduced the risk of using an arbitrary starting geometry. This task selected the lowest-energy structure from 20 sampled conformers.

Can a team repeat this work in Mira?

Yes. Mira keeps the task question, molecular input, workflow, output files, and review conclusion in the project. A team can reuse the protocol for control compounds and compare the results in the same context.