Quimb

Official Resources

• Homepage: https://quimb.readthedocs.io/
• Repository: https://github.com/jcmgray/quimb
• License: Apache License 2.0
• PyPI: https://pypi.org/project/quimb/

Overview

Quimb is a standalone Python library for quantum information and many-body calculations. It aims to bridge the gap between "easy" and "fast" by providing a high-level, interactive API for defining quantum states and operators, coupled with a powerful tensor network engine (quimb.tensor). Quimb handles arbitrary tensor network geometries (1D, 2D, 3D, hyperbolic) and integrates with advanced contraction path optimizers and hardware-accelerated backends (like CuPy, JAX, and Torch) to perform simulations of large-scale quantum systems and quantum circuits efficiently.

Scientific domain: Quantum Information, Many-Body Physics, Quantum Circuit Simulation. Target user community: Researchers in quantum computing and condensed matter physics needing rapid prototyping and flexible geometry support.

Theoretical Methods

• Tensor Network Algorithms: DMRG, TEBD, simple-update PEPS, HOTRG.
• Quantum Information: Entanglement entropy, negativity, fidelity, partial trace.
• Dynamics: Real-time evolution using TEBD, TDVP (via extensions).
• Quantum Circuits: Circuit simulation via tensor network contraction.

Capabilities (CRITICAL)

• Geometry Agnostic: Natively handles MPS, PEPS, MERA, and random tensor graphs.
• Advanced Contraction: Uses cotengra (COntraction Tree ENGRAne) to find optimal contraction paths for complex networks.
• Backend Acceleration: Can dispatch tensor operations to GPU-enabled backends (JAX, TensorFlow, PyTorch).
• Visualization: Excellent drawing tools for tensor networks, showing bond dimensions and connectivity.
• Interactive: Designed for Jupyter notebooks with rich visual outputs.
• Exact Diagonalization: Includes tools for full Hilbert space caching and ED.

Key Strengths

Flexibility

• Not restricted to lattices; can handle random graphs or quantum circuits with arbitrary connectivity.

Performance

• "Easy but Fast": Uses Numba for critical sections.
• Supports distributed computing via ray or dask.

Visualization

• Best-in-class visualization for tensor networks, greatly aiding debugging and intuition.

Inputs & Outputs

• Input: Quantum circuits (qasm), Hamiltonians, Tensor definitions.
• Output: Entanglement measures, Expectation values, Optimized states, Visualizations (.png, .svg).

Interfaces & Ecosystem

• Python: Pure Python installation.
• Integration: Works with cotengra for path finding, slepc4py for eigensolvers.
• Backends: Compatible with numpy, jax, torch, tensorflow, cupy.

Advanced Features

• Quantum Circuit Simulation: Simulates deep circuits by optimizing contraction order.
• Hyper-optimization: Can optimize contraction trees for repeated execution.
• Symmetries: Basic support for symmetries.

Performance Characteristics

• Contraction: State-of-the-art contraction path finding makes it competitive for "hard" networks (e.g., Sycamore circuits).
• Speed: GPU backends provide massive speedups for large bond dimensions.

Comparisons with Other Codes

• vs TeNPy: TeNPy is more specialized for 1D/2D ground states with U(1) symmetry and physics-heavy features. Quimb is better for random circuits, visualization, and geometric flexibility.
• vs TensorNetwork: Quimb is a higher-level framework that can use TensorNetwork as a backend. Quimb provides the physics/QI logic, TN provides the contraction engine.
• vs Qiskit: Quimb simulates circuits using tensor networks, often scaling better for low-entanglement circuits than state-vector simulators.

Application Areas

• Quantum Chaos: Studying entanglement growth in random circuits.
• NISQ Simulation: Simulating Google/IBM quantum supremacy circuits.
• Tensor Network Machine Learning: Training generative models.

Best Practices

• Path Optimization: Always pre-optimize contraction paths for large networks (reuse the path).
• Backends: Switch to jax or cupy for GPU execution.
• Visualization: Use .draw() frequently to inspect network structure.

Verification & Sources

Primary sources:

1. Repository: https://github.com/jcmgray/quimb
2. Documentation: https://quimb.readthedocs.io/
3. Papers citing Quimb (e.g., Gray & Kourtis, arXiv:2002.01935).

Confidence: VERIFIED - Active development and widely used. Verification status: ✅ VERIFIED