QMC@Home

QMC@Home
QMC@Home screensaver showing a molecular calculation
Project
Status	Completed
Category	Chemistry
Compute	CPU
Development
Developer	Martin Korth
Author	Martin Korth, Arne Lüchow, Stefan Grimme
Sponsor	University of Münster; later Ulm University
Initial release	23 May 2006
Completed	c. 2014
Software
Operating system	Windows, Linux, macOS
Metadata
Website	http://qah.uni-muenster.de/

QMC@Home (Quantum Monte Carlo at Home) was a volunteer computing project running on the BOINC platform, dedicated to developing and applying Quantum Monte Carlo (QMC) methods for use in quantum chemistry. Originally hosted at the University of Münster, Germany, with participation from the Cavendish Laboratory at the University of Cambridge, the project allowed volunteers worldwide to donate idle computing cycles to help calculate the electronic structure and molecular geometry of small molecules. The project began beta testing on 23 May 2006 and ran until approximately 2014, leaving behind a meaningful scientific legacy in computational chemistry.

Background and motivation

Interactions between molecules govern nearly every process in chemistry and biology, from the folding of proteins to the stacking of DNA base pairs. In principle, quantum chemistry can predict these interactions precisely by solving the electronic Schrödinger equation:

$\hat{H}\mathrm{\Psi }=E\mathrm{\Psi }$

where $\hat{H}$ is the Hamiltonian operator, $\mathrm{\Psi }$ is the many-electron wavefunction, and $E$ is the total energy of the system. However, exact solutions are computationally intractable for all but the smallest atoms. For decades, chemists relied on approximate methods such as DFT and CCSD(T), the latter often called the "gold standard" of quantum chemistry.

Quantum Monte Carlo methods offer a complementary route. Rather than expanding the wavefunction in a basis set, QMC uses stochastic sampling to evaluate integrals in a high-dimensional space, enabling in principle a more systematic approach to the exact solution. The key variant used by QMC@Home was Diffusion Monte Carlo (DMC), in which an ensemble of random walkers propagates through electron configuration space in imaginary time, converging toward the electronic ground state.

DMC scales favorably with system size, roughly as ${𝒪}(N^3)$ for $N$ electrons, compared to the ${𝒪}(N^7)$ of CCSD(T). This favorable scaling makes QMC an attractive target for parallelization, and crucially for the concept of volunteer computing: each DMC "walker" trajectory is essentially independent, meaning the problem can be split into thousands of separate workunits with minimal communication overhead.

Stefan Grimme, a chemistry professor at the University of Münster, and his graduate student Martin Korth recognized this potential. According to Korth, the idea of a SETI@home-style project for quantum chemistry had circulated half-jokingly in the group for some time before Korth discovered BOINC and set the project in motion.

Project history

Origins at Münster (2006)

QMC@Home entered alpha testing in early 2006 and opened public beta on 23 May 2006, hosted at qah.uni-muenster.de. The project's founding team comprised Martin Korth, Prof. Arne Lüchow (theoretical chemistry, quantum Monte Carlo expert), and Prof. Stefan Grimme (computational chemistry, specialist in noncovalent interactions). The Cavendish Laboratory at Cambridge contributed expertise in trial wavefunction construction, a key ingredient of DMC calculations.

The project was listed on the official BOINC projects page at Berkeley and rapidly grew an international volunteer community. By February 2010, QMC@Home had roughly 7,500 active participants from 102 countries, collectively contributing approximately 5 teraFLOPS of computing power. A wider estimate of approximately 14,000 volunteers was reported at the time the primary S22 benchmark calculations were completed.

The S22 benchmark campaign

The central scientific campaign of the Münster phase focused on the S22 benchmark set, a standard collection of 22 noncovalently bound molecular dimers compiled by Hobza and colleagues. The set covers three classes of noncovalent interaction:

• Hydrogen-bonded complexes: ammonia dimer, water dimer, formic acid dimer, formamide dimer, uracil dimer (HB), 2-pyridoxine/2-aminopyridine, and adenine–thymine (Watson–Crick)
• Dispersion-dominated complexes: methane dimer, ethene dimer, benzene/methane, stacked benzene dimer, stacked pyrazine dimer, stacked uracil dimer, and stacked indole/benzene
• Mixed-character complexes: ethene/ethyne, benzene/water, benzene/ammonia, benzene/HCN, T-shaped benzene dimer, T-shaped indole/benzene, and the phenol dimer; additionally adenine–thymine (stacked) and phenol dimer

Workunits distributed to volunteers handled individual DMC energy evaluations for these 22 dimers and their constituent monomers. The interaction energy of each complex is defined as:

$E_{\mathrm{i}\mathrm{n}\mathrm{t}}=E_{\mathrm{d}\mathrm{i}\mathrm{m}\mathrm{e}\mathrm{r}}-E_{\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}\mathrm{A}}-E_{\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}\mathrm{B}}$

Because interaction energies are typically small (a few kcal mol^-1) relative to the total electronic energies involved (hundreds or thousands of Hartree), achieving sub-kcal mol^-1 accuracy requires enormous statistical precision from the stochastic DMC algorithm, which in turn requires very large numbers of Monte Carlo samples — exactly the kind of computation that volunteer computing can supply.

The resulting 2008 paper in the Journal of Physical Chemistry A reported worldwide distributed QMC calculations approaching the exact solution of the electronic Schrödinger equation for noncovalent molecular interactions, providing an independent benchmark for the S22 set using DMC.

A new application: QASINO

The QMC@Home screensaver visualizing an adenine–thymine base pair calculation, one of the project's core research targets.

In March 2010, the project deployed a new application called QASINO, expanding the range of calculations available to volunteers. The project also began exploring Density functional theory as a parallel computational approach.

Relocation and rebrand (2012–2013)

By late 2012, Martin Korth had moved his research group from Münster to the University of Ulm (Institute for Theoretical Chemistry). In September 2012, the project announced it would migrate from qah.uni-muenster.de to a new server, with work planned to restart under a new URL at qmcathome.org around late October 2012.

The transition was significantly delayed. The original Münster server went offline in October 2012 and the new site was not operational until July 2013, leaving volunteers without work for several months. When the project resumed under qmcathome.org, it featured four applications:

1. Three applications related to Quantum Medicinal Chemistry (abbreviated QMC in the new context)
2. One called Clean Mobility (cleanmobility.now), analyzing new materials for electric vehicle batteries

The rebranded project expanded QMC from "Quantum Monte Carlo" toward also encompassing "Quantum Medicinal Chemistry," reflecting Korth's broadening research interests at Ulm. The Clean Mobility application attracted attention — and some frustration — from volunteers, as its workunits could run for 50 to over 100 hours with infrequent checkpointing. Korth acknowledged the issue in correspondence with a volunteer, noting that average runtimes were roughly one day and that improved checkpointing was being worked on.

Final phase and closure

The project's forums were never fully restored after the 2013 migration, and statistics exports ceased, making activity tracking difficult. The project gradually wound down around 2014 as Korth's group shifted focus toward publication and new research directions. The qmcathome.org domain was subsequently acquired by an unrelated Italian publication. The BOINC project is now listed among completed projects.

Scientific methodology

Diffusion Monte Carlo

The primary algorithm used by QMC@Home was fixed-node Diffusion Monte Carlo (FN-DMC). In this method, the ground-state wavefunction ${\mathrm{\Psi }}_0$ is projected out from a trial wavefunction ${\mathrm{\Psi }}_T$ using imaginary-time propagation:

${\mathrm{\Psi }}_0\approx \underset{\tau \to \mathrm{\infty }}{\lim }{e}^{-\tau (\hat{H}-E_T)}{\mathrm{\Psi }}_T$

where $\tau$ is imaginary time and $E_T$ is an energy offset. The fixed-node approximation enforces the correct fermionic antisymmetry by requiring walkers to respect the nodal surface of ${\mathrm{\Psi }}_T$, typically obtained from a Hartree–Fock or DFT single Slater determinant. This introduces the fixed-node error, which the research team identified as the primary remaining methodological bottleneck after the S22 calculations were complete.

Trial wavefunctions in QMC@Home used the Slater-Jastrow form:

${\mathrm{\Psi }}_T=D_{\uparrow }{D}_{\downarrow }\exp (J)$

where $D_{\uparrow }$ and $D_{\downarrow }$ are Slater determinants for spin-up and spin-down electrons, and $J$ is the Jastrow factor explicitly including electron–electron and electron–nucleus correlation terms.

Workunit design

Each workunit represented a set of DMC energy evaluations for a specific molecular geometry. The target runtime was designed to be between 4 and 48 hours on a 2.4 GHz computer, balancing scientific efficiency with volunteer computing practical constraints. The statistical nature of DMC meant that many independent runs (from many volunteers) had to be averaged to achieve the required precision in the final interaction energies.

Molecules studied

The S22 benchmark system molecules studied by QMC@Home volunteers included dimers of the following species:

System	Dominant interaction type
Ammonia dimer (1)	Hydrogen bonding
Water dimer (2)	Hydrogen bonding
Formic acid dimer (3)	Hydrogen bonding
Formamide dimer (4)	Hydrogen bonding
Uracil dimer, HB (5)	Hydrogen bonding
2-Pyridoxine / 2-aminopyridine (6)	Hydrogen bonding
Adenine / thymine WC (7)	Hydrogen bonding
Methane dimer (8)	Dispersion
Ethene dimer (9)	Dispersion
Benzene / methane (10)	Dispersion
Benzene dimer T-shape (11)	Dispersion
Pyrazine dimer (12)	Dispersion
Uracil dimer stacked (13)	Dispersion
Indole / benzene stacked (14)	Dispersion
Adenine / thymine stacked (15)	Dispersion
Ethene / ethyne (16)	Mixed
Benzene / water (17)	Mixed
Benzene / ammonia (18)	Mixed
Benzene / HCN (19)	Mixed
Benzene dimer parallel (20)	Mixed
Indole / benzene T-shape (21)	Mixed
Phenol dimer (22)	Mixed

Publications

The primary peer-reviewed publication arising from QMC@Home's volunteer computing work is:

• (2008).Toward the Exact Solution of the Electronic Schrödinger Equation for Noncovalent Molecular Interactions: Worldwide Distributed Quantum Monte Carlo Calculations. Journal of Physical Chemistry A. pp. 2104–2109. DOI: 10.1021/jp077592t.

This paper has continued to be cited in subsequent benchmark studies of DMC for noncovalent interactions, including recent work on basis set incompleteness errors in fixed-node DMC.

Legacy and context

QMC@Home was one of several quantum chemistry and molecular simulation BOINC projects of the mid-2000s, alongside Rosetta@home (protein structure prediction), Predictor@home, eOn (materials science), and AQUA@home (quantum adiabatic computing). Its contribution was particularly valuable in demonstrating that stochastic many-body electronic structure calculations — which require enormous numbers of independent samples — are exceptionally well-suited to volunteer computing architectures.

The fixed-node error problem that the team identified at the end of the S22 campaign remains an active area of computational chemistry research as of 2026, with ongoing work on multi-determinant trial wavefunctions and nodal surface optimization. The Korth group's paper continues to serve as a reference benchmark for new DMC approaches applied to noncovalent interactions.

The project was profiled in Chemical and Engineering News in 2007 as an example of how BOINC-enabled volunteer computing was enabling ambitious quantum chemistry computations that would have been impractical on a single research group's cluster.

See also

• BOINC
• SETI@home
• Rosetta@home
• Cosmology@Home
• Quantum Monte Carlo
• Density functional theory
• BOINC projects

External links

• Original project website (qah.uni-muenster.de) – Internet Archive
• Relaunched project website (qmcathome.org) – Internet Archive
• QMC@Home – Wikipedia
• Korth, Lüchow, Grimme (2008) – primary paper (DOI)