Szczepanik Research Group

Department of Theoretical Chemistry
Faculty of Chemistry, Jagiellonian University
Gronostajowa 2, 30-387 Krakow, Poland
Tel: (+48) 12 686 23 90
E-mail: dariusz.szczepanik@uj.edu.pl

◼ Download

The runEDDB program is available free of charge as precompiled binaries for Linux, Windows, and macOS — native builds for every platform on both architectures. Choose your platform below to download the latest release, matching the build to your CPU: x86_64 for Intel / AMD machines (including Intel Macs) or arm64 for Apple Silicon Macs (M1-5), Windows on ARM devices, ARM servers, and Raspberry Pi 4+. Every package is a single self-contained executable — no runtime libraries, no emulation layer, nothing else to install. See the Installation section further down this page for setup instructions.

◼ Installation

Each platform package contains the runEDDB executable, an automatic installer (install.sh on Linux / macOS, install.bat on Windows), and the third-party attribution and license files (THIRD-PARTY-LICENSES.txt together with the verbatim LGPL-3.0, GPL-3.0 and Apache-2.0 texts covering the bundled GFN2-xTB / DFT-D4 components). The installer copies runEDDB to a permanent ~/runEDDB/ folder and adds it to your PATH — no administrator privileges required. To update later, just download the new release and re-run the installer.

Which architecture do I have? On Linux and macOS run uname -m — it prints x86_64 or aarch64 / arm64. On Windows open Settings → System → About and read System type. In the steps below, replace ARCH with x86_64 or arm64 accordingly.

1. Linux

x86_64 — Intel / AMD 64-bit machines, the most common desktop and cluster case.  arm64 — ARM servers (AWS Graviton, Ampere), Raspberry Pi 4+ and Apple-silicon Linux guests.

1. Unzip runEDDB_linux_ARCH.zip (from your file manager or via unzip runEDDB_linux_ARCH.zip).
2. Enter the folder and run the installer: cd runEDDB_linux_ARCH && bash install.sh.
3. Open a new terminal window (or run source ~/.bashrc in the current one).
4. Verify: runEDDB --help.

2. Windows

x86_64 — Intel / AMD machines.  arm64 — Windows on ARM devices (Snapdragon X and similar) and Windows 11 ARM64 running in a VM on Apple Silicon. Both are native builds: no x64 emulation and no WSL are needed on ARM hardware.

1. Right-click runEDDB_windows_ARCH.zip → Extract All… → Extract.
2. Open the extracted folder and double-click install.bat.
3. Open a new PowerShell or CMD window (press Win+R, type powershell, Enter).
4. Verify: runEDDB --help.

3. macOS

arm64 — Apple Silicon Macs (M1-5).  x86_64 — Intel Macs.

1. Double-click runEDDB_macos_ARCH.zip to extract.
2. In Terminal (Cmd+Space → type Terminal), enter the folder and run the installer: cd ~/Downloads/runEDDB_macos_ARCH && bash install.sh. The installer automatically removes the macOS quarantine flag.
3. Open a new Terminal window (or run source ~/.zshrc).
4. Verify: runEDDB --help.

◼ Changelog

Compilation 2026-Aug-15 (Current)

• Added --nao-algorithm cc-layers: correlation-consistent layered NAO construction (ccNAO) with monotonic radial occupations.
• Added --table shell,subshell options (shell/subshell-resolution natural populations).
• Fixed dropped Rydberg NAOs (e.g. 3d) in the rigorous algorithm at large basis sets.
• Fixed/added angular-momentum support up to k functions (L = 7, e.g. cc-pV7Z) across all wavefunction input formats.
• Fixed NAO labels with principal quantum numbers ≥ 10 (previously printed as * in the --table ao output).
• Fixed empty AO/NPA tables and exports for single atoms and non-bonded atoms in the global (EDDB_G) scope.

Compilation 2026-Aug-01

• Added native Windows on ARM64 binaries (self-contained, no x64 emulation or WSL required).
• Improved the precision of the DFT-D4 dispersion reference data in the spGFN2-xTB engine (Sr and Ba).
• Added --heavy-atoms (-h) for the EDDB_H scope, which works like EDDB_G with all hydrogen atoms excluded from BOP.
• Added --cube-align and --cube-orb2dens options (see manual for more details).
• Fixed and extended --cube-ref option (now it works with both orbitals and densities).
• Fixed the cumulative-idempotent quasi-degenerate layer construction for post-HF BOP calculations.
• Connectivity mask (EDDB_G, EDDB_H,EDDB_F and EDDB_R) is now always determined in the full NAO basis.
• Added --wbo-thresh X (-w X) to set the Wiberg cutoff for the BOP pair gate (default 0.001).
• Added --xtb-frac-occ (diagnostic flag): keeps the Fermi-smeared fractional occupations in the GFN2-xTB calculations.

Compilation 2026-Jul-12

• Fixed --export-fchk options involving geometry-input files.
• Added --nao-algorithm frozen-nmb|rigorous option (frozen-nmb is the default).
• Added --export-cube ao|no|nobd|bo|xtb[LIST] options to export LIST of orbitals directly to Gaussian cube files.
• The Galerkin projection of exported NAOs is no longer applied silently (it can be enabled via the --export-lproj flag).

Compilation 2026-Jul-01

• New UTF-8 interface (--ascii forces the ASCII mode).
• The --output-* options replaced by --table, --export-fchk, --export-cube.
• Added --cite option (references are emitted per stage and per option).

Compilation 2026-Jun-04

• Added --output-d{ed,eddb,edlb,bdf} options: differential maps (ΔED / ΔEDDB / ΔEDLB / ΔBDF) against a reference cube.
• Added --output-moa option: the ED / EDLB / EDDB populations in the MO resolution.
• Added new geometry input formats: MDL Molfile (.mol / .sdf), Tripos MOL2 (.mol2) and Protein Data Bank (.pdb).
• Added --xyz-frame option: pick a structure from a multi-frame .xyz / .sdf / .mol2 / .pdb trajectory.
• Added --xyz-center and --xyz-translate options to recenter or rigidly translate the input geometry.
• Added --xyz-charge and --xyz-mult options to override charge and multiplicity for geometry inputs.

Compilation 2026-Jun-01

• Added native spGFN2-xTB engine: electron-delocalization (EDDB/BDF) maps straight from a plain .xyz geometry (e.g. X-ray).
• Added --xtb-etemp/conv/maxiter/noaes/nospin options to control Self-Consistent Charge (SCC) convergence.
• Added --output-xtb option to export the converged spGFN2-xTB wavefunction to a Gaussian .fchk file (for visualization).
• Added new diradical / multireference diagnostics to --output-no option.
• Fixed --output-ao consistency with --auto-pi and --select-nobd options.

Compilation 2026-May-21

• Added --output-no option: exports natural orbitals (NOs) of the spinless 1-electron density to a Gaussian .fchk file.
• Added Cor / Val / Ryd breakdown of each NO printed alongside the NO occupations.
• Added simple multireference-character diagnostics (Mayer, Head-Gordon, Yamaguchi) under the NO table.
• All FCHK exports now embed orbital occupations so Avogadro2 displays them directly in the orbital-energy column.
• Fixed --auto-pi for topologically differentiated π-conjugated systems.

Compilation 2026-May-11

• Added Turbomole (HF/KS) and GAMESS-US (HF/KS, post-HF) support.
• Added native ECP support across every parser .
• Fixed NAO principal-quantum-number labelling.
• Added AO overlap normalization sanity check runs in every parser.
• Improved autodetection of π-NOBDs (--auto-pi option).
• Fixed per-primitive normalization bug for cc-pVnZ (n=6-8).
• Fixed EDDB_F and EDDB_P summary.

Compilation 2026-May-01

• Reads Gaussian .fchk / .fch, ORCA .json, and Multiwfn / ORCA / PySCF .molden directly — NBO7 is no longer required.
• In-house NAO driver — improved and parallelized for multi-core machines; numerically matches NBO7 to sub-percent.
• Post-HF correlated densities (MP2, CCSD, CCSD(T), CASSCF, CI) are auto-detected and used in the analysis.
• Default analysis basis switched to NVB (valence only); NMB and full NAO remain available.
• New --basis VALUE (or -b) flag replaces the legacy --nao / --nmb / --nvb triplet (case-insensitive).
• Redesigned --cube-size N: the sign and zero of N encode four modes — scope-aware grid, full-molecule grid, skip cubes, or clone exact grid from a reference .cube file.
• Cube clone mode (-c 0 ref.cube) makes differential cube maps trivial — e.g. EDDB(method A) − EDDB(method B).
• Three-tier verbosity: default (stage messages + atomic table + summary), true-quiet -q (banner + summary only), verbose -v (all three tables + detail blocks).
• Removed: --cube-full (use -c -N instead).
• Tutorial and Manual rewritten for the new pipeline; C₇H₇FeCl walkthrough now uses an ORCA .molden input.

Compilation 2026-Apr-11

• Reads NBO7 .49 archives (the only input format).
• NMB (Natural Minimal Basis) as the default analysis subset; full NAO available.
• EDDB / EDLB / BDF / ED density export to Gaussian .cube files.
• NOBD orbitals and NAOs export to Gaussian .fchk files.
• Global / Fragment / Pathway analysis scopes (-g / -f / -p).
• Automatic π-NOBD selection (--auto-pi) and manual NOBD dissection (--select-nobd).
• Cyclic Delocalization Index (CDI) printed automatically in pathway mode.
• Tunable Wiberg bond-order threshold (--wbo-thresh).
• Multi-core parallel cube evaluation (--ncores).
• Linux, Windows, and macOS binaries.