deal.II

deal.II
Name	deal.II
Programming language	C++
Operating system	Cross-platform
Genre	Finite element library
License	LGPL

deal.II

deal.II is an open-source finite element library implemented in C++ that supports adaptive mesh refinement, parallel computing, and a wide spectrum of numerical methods for partial differential equations. It is used by researchers and engineers at universities, national laboratories, and companies worldwide for simulations in mechanics, fluid dynamics, electromagnetics, and geophysics. The project integrates with external packages for linear algebra, meshing, and visualization and has influenced numerical software ecosystems and education in computational science.

Overview

deal.II provides a comprehensive toolkit for solving boundary value problems using the finite element method, supporting higher-order elements, hp-adaptivity, and multigrid techniques. The library is designed for extensibility and interoperability with packages such as PETSc, Trilinos, METIS, ParMETIS, and SLEPc, enabling large-scale simulations on clusters operated by institutions like Lawrence Berkeley National Laboratory, Sandia National Laboratories, and Oak Ridge National Laboratory. Users employ deal.II in conjunction with mesh generators like Gmsh and Netgen, visualization tools such as ParaView and VisIt, and algebraic libraries including Eigen and MKL.

Features and Capabilities

deal.II implements support for continuous and discontinuous Galerkin methods, mixed finite elements, and stabilized formulations, with built-in routines for adaptivity, error estimation, and hp-refinement. The library includes interfaces to linear solvers and preconditioners from PETSc, Trilinos, and Hypre, and supports direct solvers like MUMPS and SuperLU, enabling coupling with performance libraries such as OpenBLAS and Intel MKL. Parallel capabilities rely on Message Passing Interface implementations like OpenMPI and MPICH, while I/O and checkpointing leverage HDF5 and SIONlib for workflows common in research at institutions including CERN, NASA, and EURATOM.

Architecture and Design

deal.II features a layered design separating discretization, linear algebra, mesh management, and I/O, facilitating integration with third-party packages like p4est for scalable octree meshes and Zoltan for dynamic load balancing. The software uses modern C++ idioms and template metaprogramming, and exposes abstractions for DoFHandler, FEValues, and ConstraintMatrix to represent finite element spaces and constraints used in methods developed at MIT, Stanford, and ETH Zurich. The design enables implementations of multigrid hierarchies, geometric and algebraic multigrid, and matrix-free operator evaluation as used in high-order codes at Caltech and Imperial College London.

Development History

The project originated from academic groups and has evolved through contributions from researchers associated with institutions like the University of Heidelberg, Technical University of Munich, and Lawrence Livermore National Laboratory. Over time, deal.II incorporated capabilities from collaborations with software projects such as DUNE, FEniCS, and FreeFEM, and adopted practices from open-source communities similar to KDE and Apache Software Foundation. Releases have reflected advances in HPC, with work targeting exascale initiatives led by DOE, European Commission programs like Horizon 2020, and national computing centers including PRACE and Compute Canada.

Applications and Use Cases

deal.II is applied across engineering and physical sciences for problems in solid mechanics, fluid mechanics, electromagnetics, and poromechanics, with example studies at institutions such as Stanford University, University of Cambridge, TU Delft, and École Polytechnique. Industrial adopters include Siemens, General Electric, and Rolls-Royce for design analysis, while geophysical and climate groups at NOAA, ECMWF, and the British Antarctic Survey use deal.II for subsurface flow and ice-sheet modeling. Multiphysics couplings enable research in biomechanics at Harvard Medical School and material science at Max Planck Institute and Oak Ridge, while academic courses at Princeton University and University of Oxford use deal.II in computational science curricula.

Performance and Benchmarks

Performance studies compare deal.II workflows against benchmarks using Trilinos, PETSc, and Hypre on supercomputers at Argonne National Laboratory, National Center for Supercomputing Applications, and Lawrence Livermore National Laboratory. Matrix-free implementations and sum-factorization techniques demonstrate scalability in high-order simulations similar to results reported by groups at RWTH Aachen, KAUST, and Leibniz Supercomputing Centre. Profiling and performance engineering leverage tools from Intel VTune, GNU gprof, and Allinea MAP, and optimization targets BLAS/LAPACK implementations, GPUs via CUDA and HIP, and task-based runtimes such as Intel TBB and OpenMP used at NVIDIA, AMD, and Cray.

Licensing and Community

deal.II is distributed under a permissive LGPL-style license and is developed by a community of contributors from universities, national labs, and commercial entities including Microsoft Research and IBM Research. The project maintains user and developer forums, mailing lists, and workshops attended by participants from institutions like ETH Zurich, University of Bonn, and TU Darmstadt; it collaborates with standardization bodies and funding agencies such as NSF, EPSRC, and DFG. Documentation, tutorials, and example codes support adoption in research groups at Columbia University, University of Michigan, and Kyoto University, while community governance follows open-source practices akin to those of the Linux Foundation and Apache projects.

Category:Finite element software