Hosted Quantum Chemistry
Run PySCF online, no installation
Run PySCF 2.14.0 directly in your browser without local Python environments, compilation hurdles, or cluster maintenance. ChemDFT executes genuine PySCF jobs in isolated cloud containers with verifiable cryptographic provenance.
Sign in with Google. No credit card required. Free for researchers during the beta.
What you can run, and the limits
ChemDFT provides hosted cloud compute for quantum chemistry and materials modeling within strictly defined, transparent boundaries. Calculations execute in isolated Linux worker environments on Amazon Web Services (AWS) pilot hardware, automatically provisioning on demand and shutting down after approximately fifteen minutes of idle time. Because computational resources are shared and free during the beta, hosted system size limits are modest and strictly enforced before submission.
For molecular quantum chemistry using PySCF 2.14.0, the hosted size limit is 12 atoms for geometry optimisation and molecular orbital calculations, and 6 atoms for harmonic vibrational frequencies and time-dependent DFT (TDDFT). For semi-empirical calculations using xTB 6.7.1, the hosted limit is 50 atoms. For conformational search using CREST 3.0.2, the hosted limit is 12 atoms for neutral singlet hydrocarbons and oxygenates. For periodic solid-state calculations using Quantum ESPRESSO 7.6 or periodic PySCF, the unit cell limit is 20 atoms.
ChemDFT does not replace high-performance computing clusters or supercomputers for large macromolecules, heavy polymers, or broad crystal supercells. Every result is provided as a transparent research aid that researchers must independently check and verify before publication. The artificial intelligence assistant plans workflows and explains findings, but it never writes, interpolates, or modifies a numerical calculation result.
Hosted PySCF calculations are bounded by strict size caps: 12 atoms for ground-state energy, geometry optimization, and molecular orbitals; and 6 atoms for harmonic vibrational frequencies and time-dependent DFT (TDDFT).
What is PySCF?
The Python Simulations of Chemistry Framework (PySCF) is a state-of-the-art open-source quantum chemistry and solid-state physics codebase developed by an international consortium of computational chemists led by Qiming Sun and Garnet Kin-Lic Chan. Built with Python for modular workflow design and C/C++ libraries (such as libcint and Libxc) for high-performance numerical integrals, PySCF has become one of the most powerful and widely cited modern computational chemistry engines.
PySCF excels in ground-state Hartree–Fock, Kohn–Sham density functional theory, molecular geometry optimization, vibrational frequencies, excited-state time-dependent DFT, and advanced electronic structure methods. While PySCF is freely available under the BSD-3-Clause open-source license, running it typically requires installing a complete Python science stack, configuring specialized C-extension dependencies, and managing local compute resources.
Hosted PySCF 2.14.0 capabilities in ChemDFT
ChemDFT hosts PySCF version 2.14.0 in standardized, reproducible Linux containers. Within ChemDFT, you can execute the following molecular electronic structure tasks:
1. Electronic Methods and Functionals: Ground-state Restricted Hartree–Fock (RHF), Unrestricted Hartree–Fock (UHF), the hybrid B3LYP functional (B3LYP), and the generalized gradient approximation PBE functional (PBE).
2. Standard Gaussian Basis Sets: Minimal STO-3G for rapid instructional demonstrations; Pople split-valence 6-31G for foundational studies; and Ahlrichs balanced def2-SVP and diffuse def2-SVPD basis sets for rigorous molecular geometry and orbital analysis.
3. Calculation Tasks: Single-point SCF electronic energy, equilibrium geometry relaxation using PySCF analytic gradients, harmonic vibrational frequencies and normal mode displacement vectors, canonical molecular orbital eigenvalues and occupations, Time-Dependent DFT (TDDFT) singlet excited states, and gas-phase thermochemical properties (enthalpy, entropy, and Gibbs free energy).
On-demand cloud worker infrastructure
ChemDFT does not run Python or PySCF inside your local web browser. Browser-based WebAssembly implementations of heavy quantum chemistry engines suffer from severe memory constraints, lack optimized linear algebra libraries (BLAS/LAPACK), and struggle with multi-threading.
Instead, ChemDFT connects your browser to dedicated AWS compute instances running hardened Linux worker containers. When you submit an approved calculation, an on-demand worker spins up, retrieves the normalized calculation parameters, executes PySCF with optimized BLAS libraries, and generates the raw output and checkpoint files. To optimize research funding and operational costs, the compute worker remains active during continuous use and automatically powers off after approximately fifteen minutes of idle time. Subsequent jobs submitted while the worker is asleep will experience a brief initialization window while the container boots.
Molecular geometry optimization and geomeTRIC note
For molecular geometry optimization, ChemDFT couples PySCF electronic energy gradients with the geomeTRIC 1.1.1 coordinate optimizer developed by Lee-Ping Wang. geomeTRIC utilizes translation-rotation internal coordinates (TRIC) to achieve robust, rapid convergence to local minima on the potential energy surface. As noted across the ChemDFT site, geomeTRIC optimizes molecular geometries using PySCF gradients; periodic crystal structures are not optimized with it.
Hosted geometry optimization is enabled under explicit owner decision. The geomeTRIC 1.1.1 license includes a non-AI clause (clause 4), and whether this clause applies to general platform execution remains an open legal decision. ChemDFT maintains complete transparency regarding software licensing across its entire stack.
Scientific integrity and immutable provenance
In ChemDFT, every PySCF job is cryptographically bounded. The normalized input file is hashed with SHA-256 before submission. Upon completion, the raw engine logs, total energies, convergence iterations, and orbital coefficients are validated by the Scientific Result Firewall. If an SCF calculation fails to converge within the allotted iteration limit (default 100 cycles), the result is flagged as unconverged and refused publication as authoritative data.
AI models assist you with planning basis sets and troubleshooting convergence errors, but they cannot edit or author calculation numbers. Every value you see in the interface is read directly from genuine PySCF output files.
Frequently Asked Questions
Do I need to install Python, conda, or PySCF on my computer?
No. You do not need to install Python, PySCF, or any compilers. All calculations run on ChemDFT's cloud compute workers, and you interact with the system entirely through your web browser.
How long does a PySCF job take to run on ChemDFT?
Small single-point calculations (such as water or benzene with minimal basis sets) complete in seconds. If the cloud worker is currently powered off, allow approximately 1 to 2 minutes for container initialization.
What basis sets are supported for PySCF calculations?
ChemDFT currently supports STO-3G, 6-31G, def2-SVP, and def2-SVPD in gas-phase molecular calculations. Additional basis sets are evaluated in accordance with our hosted validation roadmap.
Can I download the raw PySCF output and log files?
Yes. Every calculation preserves its raw engine log files, input parameters, and convergence records, which can be inspected directly in the workspace or exported for publication.