NanoHive@Home

NanoHive@Home
NanoHive@Home project logo
The NanoHive@Home screensaver, showing the work unit identifier, host and total credit, a progress bar, and the running quantum chemical optimization log
Project
Status	Completed
Category	Nanotechnology (computational chemistry, molecular nanotechnology)
Compute	CPU
Requires	None (the client bundled the NanoHive-1 simulator, MPQC libraries and the graphics/screensaver executable)
Development
Developer	Brian Helfrich (NanoHive@Home and NanoHive-1); Nanorex, Inc.
Author	Brian Helfrich
Sponsor	Nanorex, Inc., Bloomfield Hills, Michigan
Maintainer	Brian Helfrich (project engineer); Dr. Damian G. Allis (project scientist)
Initial release	October 4, 2006  (20 years ago) (open beta; closed beta from 25 May 2006)
Completed	Science run ended May 2007; project closed 2008
Discontinued	2008
Repository	https://sourceforge.net/projects/nano-hive/
Software
Written in	C, C++ (TCL simulation scripting)
Operating system	Windows (x86, amd64, x86_64); Mac and Linux clients announced but never released
Size	4 MB download, expanding to about 20 MB on disk
BOINC statistics
Stats as of	January 14, 2008  (18 years ago)
Performance	About 3 TFLOPS at peak (project statement); 4.72 TFLOPS recorded by BOINCstats on 27 April 2007
Active users	0 (project closed)
Total users	3,165
Active hosts	0 (project closed)
Total hosts	7,166
Analytics
RAC	0
Credit/day	0
GPU performance	None (no GPU application)
CPU performance	About 3 TFLOPS at peak
Metadata
Website	http://www.nanohive-1.org/atHome/ (offline; archived)
License	GNU GPL-2.0 and LGPL-2.0 (NanoHive-1 core and plugins)

NanoHive@Home (often abbreviated NHAH or NH@H) was a volunteer computing project running on the Berkeley Open Infrastructure for Network Computing (BOINC) platform between 2006 and 2008. It was created by software architect Brian Helfrich and sponsored by the molecular nanotechnology software company Nanorex, Inc., with quantum chemist Dr. Damian G. Allis serving as project scientist. The project's stated goal was "to perform large-scale nanosystems simulation and analysis that is otherwise too intensive to be calculated via normal means, and thereby enable further scientific study in the field of nanotechnology."

Unlike most BOINC projects of its era, NanoHive@Home was not run by a university. It was a not-for-profit effort attached to a commercial CAD software company, it was completely open source, and it placed all of its results in the public domain. Its single completed science campaign, the Tooltip Failure Mode Search Project, ran from February to May 2007 and produced roughly 80,000 candidate molecular geometries and 200,000 quantum chemical calculations. Those results were eventually published in 2011 in the Journal of Computational and Theoretical Nanoscience.

The project is remembered fondly in the volunteer computing community for its unusually informative screensaver, which displayed the live quantum chemistry log of the molecule being optimized. A contemporary post in the BOINCstats forum called it "the project with the most cool screensaver." Members of BOINC Synergy were active on the project, and a BOINC Synergy member was featured as NanoHive@Home's "User of the Day" in May 2007.

Scientific background

Diamondoid cage hydrocarbons: adamantane (1), diamantane (2), triamantane (3) and a tetramantane (4). The tooltips studied by NanoHive@Home were germanium-substituted polymantanes built from cages of this kind.

Diamondoid mechanosynthesis

Mechanosynthesis is the proposed fabrication of structures by mechanically positioning individual reactive molecules so that chemical bonds are formed at chosen sites, rather than by allowing reagents to meet by diffusion in solution. In molecular nanotechnology, the most-studied variant is diamondoid mechanosynthesis (DMS), in which stiff hydrocarbon cage molecules (adamantane, diamantane and larger polymantanes) act as the framework for a tooltip that carries a two-carbon feedstock unit, the carbon dimer ${\mathrm{C}}_2$, and deposits it onto a growing diamond workpiece.

The tooltips examined by NanoHive@Home came from a published survey of germanium-substituted polymantane dimer placement motifs by Robert A. Freitas Jr., Damian G. Allis and Ralph C. Merkle, and from the earlier DC10c tooltip study by Allis and K. Eric Drexler. In these designs the dimer is held between two germanium atoms, so the reactive geometry can be written schematically as $\mathrm{G}\mathrm{e}\cdots {\mathrm{C}}_2\cdots \mathrm{G}\mathrm{e}$.

Why failure modes matter

A tooltip is only useful if the geometry that performs the chemistry, the "ready" or loaded tooltip, is also the geometry the molecule prefers to sit in. If some other arrangement of the same atoms is lower in energy and reachable, the tool can quietly rearrange itself into a shape that can no longer transfer its dimer. Allis described the distinction between reactivity and stability with characteristic humour on the project site:

"A knife with its blade exposed is certainly the more functional geometry for such an object, but the knife with its blade retracted into the handle is the geometry least likely to damage the blade."

Because a manufacturing system might cycle a tooltip on the order of $10^{12}$ to $10^{23}$ times, even rare rearrangements matter. Allis pointed out that a tool surviving $10^{12}$ operations at one operation per 100 nanoseconds lasts a little under ten days, while $10^{23}$ operations at the same rate would last roughly 2.7 million years. Cataloguing the accessible defect geometries, or pathologies, of each candidate tooltip was therefore treated as a necessary last step of tooltip design, and it was exactly the kind of embarrassingly parallel search that volunteer computing does well.

The Q-SMAKAS method

Ball-and-stick model of adamantane, the smallest diamondoid cage. Tooltip frameworks in the NanoHive@Home survey were built from adamantane, diamantane, iceane and twistane cages with germanium substitutions.

The search protocol devised for the project was named Q-SMAKAS, for Quantum Search for Minimum Alternatives in Kinetically-Accessible Space. Allis contrasted it with the lower-energy conformational searching used by protein folding projects, which he labelled Q-SMACS: "Q-SMACS is to a hot water bottle what Q-SMAKAS is to a lit match."

The pipeline had three stages.

1. Reference optimization. Each tooltip was first energy-minimized with Sandia National Laboratories' Massively Parallel Quantum Chemistry program (MPQC) at the restricted Hartree-Fock RHF/3-21G level of theory. Seven of the larger tooltips (MCB5Ge, MCB57Ge, DCB75Ge, C100GeATD, C100GeATS, C100GeCTS and MCB75Ge) converged too slowly and were run at RHF/STO-3G instead.
2. Thermal deformation. Using the optimized geometry, a molecular dynamics simulation was run in GROMACS for 50,000 iterations at high temperature, with a snapshot taken every 40 to 50 iterations. The thermal energy shook each cage far away from its designed shape, generating kinetically accessible deformations.
3. Distributed re-optimization. Every snapshot became a BOINC work unit: a fresh RHF/3-21G or RHF/STO-3G geometry optimization that relaxed the deformed structure downhill into whatever local minimum it happened to fall into. Volunteers' computers performed the overwhelming majority of these optimizations.

The severity of the deformation was tracked with the root-mean-square deviation of atomic positions relative to the reference geometry,

$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{D}=\sqrt{\frac{1}{N}\sum _{i=1}^N{|{\mathbf{𝐫}}_i-{\mathbf{𝐫}}_i^0|}^2}$

where ${\mathbf{𝐫}}_i$ is the position of atom $i$ in the snapshot and ${\mathbf{𝐫}}_i^0$ its position in the reference structure. A baseline run at 1500 K produced only $\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{D}\approx 0.02$ and yielded no failure modes at all. The team therefore stepped the thermostat upward in 500 K increments, through 2000 K, 2500 K, 3000 K, 3500 K, 4000 K and 4500 K, until deformations of at least $\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{D}\approx 0.04$ were obtained. Several candidate failure modes appeared at that level. Work unit names encoded the run temperature with a letter, from z for 1500 K through t for 4500 K.

Energies were reported in hartree, the atomic unit of energy, with completely dissociated nuclei and electrons defined as the zero of energy, so that bound molecules have negative total energies. Only structures with identical atom counts computed at identical levels of theory were compared, and the quantity of interest was the relative energy

$\mathrm{\Delta }E=E_{\mathrm{d}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{c}\mathrm{t}}-E_{\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{d}\mathrm{y}},1E_{\mathrm{h}}\approx 2625.5\mathrm{k}\mathrm{J}\mathrm{m}\mathrm{o}{\mathrm{l}}^{\mathrm{-}\mathrm{1}}$

A defect with $\mathrm{\Delta }E<0$ is more stable than the working tooltip, which is the unwelcome case: the tool "wants" to be broken. The project shoutbox recorded a participant asking what the recurring note "dimer cracked" meant, and the project engineer explaining that a crack in the tooltip base which lowers the energy "is not good because then it means the tooltip naturally _wants_ to be in a broken configuration."

A crucial caveat, stated plainly by the project and repeated in the published paper, is that Q-SMAKAS locates minima but not the barriers between them. Whether a given pathology is reachable in a working device depends on the transition state energy $E_a$ and the operating temperature through a Boltzmann-type factor,

$k\propto \exp (-\frac{E_a}{k_{\mathrm{B}}T})$

so identifying accessible minima was explicitly framed as an important but separate follow-up problem.

Simulation campaigns

FineMotionController_1 (beta)

The first simulation run on the network was a beta test payload. It tested the molecular stability of the fine-motion controller for molecular assembly, a mechanism designed by K. Eric Drexler in which eight rotating rings drive struts attached to a platform, giving fine control of the platform in the x, y, z, roll, pitch and yaw degrees of freedom. The simulation was marked "over" by early 2007.

Nanofactory_1: the Tooltip Failure Mode Search Project

A molecular propeller, one of the hypothetical nanoscale machine components of the kind whose feasibility motivated projects such as NanoHive@Home. The NanoHive-1 simulator was written as a general tool for studying such structures.

The project's only full science campaign was announced on 24 January 2007, with work units queued from 29 January 2007. The news post promised that this run would behave "more like a typical BOINC project": no cap on work units per host, the whole batch queued more or less at once, and deadlines of about one week.

The target set consisted of 25 published carbon dimer tooltips, each examined in both a loaded state (L, holding the ${\mathrm{C}}_2$ feedstock and ready to deposit) and an unloaded state (U, after deposition and awaiting recharge), giving 50 structures in all. Named tooltips in the results gallery included DC10c, AdamGe22, AdamGe33, Diad3Ge22, TwistaneGe, DCBIceane7Ge, the DCB series (DCB5Ge, DCB6Ge, DCB55AGe, DCB55BGe, DCB55CGe, DCB57Ge, DCB65Ge, DCB75Ge), the MCB series (MCB5Ge, MCB55Ge, MCB57Ge, MCB75Ge) and the C100/C110/C111 germanium tooltips. Three further molecules, biotin, n-heptane and n-octane, were run as method evaluation cases to check whether the same machinery could be used at ordinary temperatures for plain conformational searching.

On 24 April 2007 the project announced that the campaign was "coming to a successful end" and that results were being organized for the Results Gallery and for peer-reviewed publication. Both the loaded and unloaded galleries were duly posted, with PDB coordinate files, energies and per-defect commentary for each structure.

HiveArena (never launched)

A third simulation, HiveArena, was listed as "in dev." but never ran. It was a plan for a public competition in which entrants would submit nanobot designs into the NanoHive-1 simulation space and compete at goals such as a speed race, a last-bot-standing melee, or a challenge to disassemble a population of hostile "virus" bots. Volunteers would have watched the contests unfold in the screensaver before anyone else, with crunching teams credited in order of work units contributed and winning teams choosing the soundtrack. The project even published a frequently asked question about how it would prevent participants from cheating by falsifying results, answering that the simulation space was subdivided very finely and that every work unit was independently checked by a second, and if necessary a third, participant.

Technology

NanoHive-1

At its core, NanoHive@Home ran the NanoHive-1 Nanospace Simulator, a modular simulator written by Brian Helfrich for modelling the physical world at nanometer scale. NanoHive-1 was designed around two problems: the sheer number of interacting factors involved in simulating nanospace, and the computational intensity of evaluating those factors for every atom at every time and space quantum. Its answer was a plugin framework in which physical interaction calculators, entity traversal schemes, data import and export formats, results management and simulation control were all replaceable modules.

Successive releases added plugins that are visible in the project's methodology, including MPQC_SClib, which wrapped the Scientific Computing toolkit of the Massively Parallel Quantum Chemistry program, BasicCellTraverser for multithreaded subdivision of the simulation space, AIREBO for the adaptive intermolecular reactive empirical bond order potential, OpenBabelImportExport for file format conversion, and SocketsControl for driving the simulator over TCP. Version 1.2.0 Beta 1, released in February 2006, was the release that made NanoHive-1 "fully distributable across computing networks." The simulator is written in C and C++ with TCL used for simulation workflow scripting, and is distributed under the GNU GPL version 2 and LGPL version 2.

BOINC integration

BOINC was bolted onto NanoHive-1 with two purpose-written plugins:

• BOINC_PIC_Control connected the master NanoHive-1 instance to the BOINC server components, generating and re-assimilating work units. "PIC" stands for Physical Interaction Calculator, which in practice means one computer running NanoHive-1, or a cluster sharing memory.
• BOINC_ClientControl connected the participant's BOINC main program to the slave NanoHive-1 instance, unpacking and launching the graphics and screensaver program and the simulator, monitoring progress, and reporting claimed credit back through the BOINC Manager.

All of the NanoHive@Home components were open source and were distributed through the SourceForge CVS repository for the nano-hive project, with anonymous pserver checkout documented on the project website.

The applications page listed two BOINC applications: the original Nano-Hive@Home application, version 5.08 for Windows/x86, installed 4 October 2006, and NHAH Quantum Search for Minimum Alternatives in Kinetically-Accessible Space, version 5.08 for Windows/x86, Windows/amd64 and Windows/x86_64, installed 22 March 2007.

Client requirements and work unit characteristics

The BOINC Manager. Volunteers attached to NanoHive@Home by entering the project URL http://www.nanohive-1.org/atHome/ , taking care with the capital "H" in "atHome".

The published system requirements covered Windows XP Professional and Home (Service Pack 1 or 2), Windows XP Tablet PC Edition, Windows Server 2003, Windows 2000 (Service Pack 2) and Windows NT (Service Pack 6 or 6a), with 128 MB of RAM (256 MB recommended) and up to 100 MB of free disk space. Mac and Linux clients were described as "in development" and a Solaris client as "planned"; none of them was ever released.

Project characteristics as published in the Getting Started page

Property	Value
Client application size	4 MB download, expanding to 20 MB on disk
Work unit size	About 10 KB, expanding to 20 KB
Result upload size	About 200 KB
Work unit processing time	40 minutes to 140 hours
Work unit deadline	9 to 12 days (about 7 days for the 2007 tooltip run)
Quorum	2 or 3, depending on the simulation

The extreme spread in run time, from under an hour to nearly six days, was not a scheduling accident. As the project explained, it is "intrinsic to the search algorithm in our quantum mechanics software": a geometry optimization takes as many steps as the potential energy surface demands, and a badly mangled 4500 K snapshot can wander for a very long time before it settles.

Credit followed a quorum rule. With two valid results, the canonical credit was the mean of the two valid claims,

$C=\frac{1}{2}(c_1+c_2)$

and both claimants received it, while other participants who had made a reasonable attempt received their own claim capped at $C$. With three or more valid results the highest and lowest claims were discarded before averaging.

The screensaver

The screensaver was the project's signature feature and doubled as a genuine progress readout. Down the left side it showed the participant's username, team, host credit and total credit. Beneath that came the work unit identifier, which was itself descriptive: a molecule name, an L or U marking a loaded or unloaded tooltip, and a deform_# suffix identifying which thermally deformed variant was being computed. The BOINC slot number, CPU time, percentage complete and a progress bar followed.

The most distinctive element was the live activity log, which exposed the two nested loops of the quantum chemical optimization. Energy iterations printed as, for example, 12.06 dE=2.024e-005 [3.337e-006], where 12 is the geometry optimization iteration, 06 the energy iteration within it, dE the achieved energy convergence

$\mathrm{d}\mathrm{E}=|E_n-E_{n-1}|$

and the bracketed figure the target accuracy. Optimization steps printed their own convergence tests on the maximum gradient and maximum displacement, each with its threshold in brackets and a plain "yes" or "no" for whether the criterion had been met. In effect volunteers watched the molecule roll downhill toward a minimum in real time.

History

Timeline of NanoHive@Home

Date	Event
25 May 2006	Closed beta testing begins with a handful of testers; account creation closed.
30 June 2006	New project website opens with information pages and discussion forums.
25 July 2006	Open enrollment delayed while a bug preventing correct work unit processing on some computers is fixed.
23 August 2006	Closed beta by a small team of BOINC experts continues; open enrollment promised for the following month.
20 September 2006	Enrollment for beta testing opens at 18:00 UTC; the project URL changes to http://www.nanohive-1.org/atHome/ and earlier closed beta testers must detach and reattach.
4 October 2006	Open beta testing begins; the Windows/x86 application version 5.08 is installed.
6 October 2006	Work unit production paused to fix bugs and to add a limited-cache mechanism improving work unit distribution and parallelism.
27 to 30 October 2006	Beta test round 5.08 / client 1.21 / server, with the new limited-cache mechanism.
24 January 2007	The Tooltip Failure Mode Search Project is announced.
29 January 2007	Work units for the tooltip run enter the queue.
11 to 12 February 2007	Roughly two hours of scheduled maintenance downtime.
17 February 2007	Client NHAH_QSMAKAS version 5.06 released, fixing a modal error dialog that blocked the BOINC Manager and correcting a progress calculation that stalled work units at 99.99 percent; the per-host daily result limit is reset to 50.
22 March 2007	Version 5.08 of the QSMAKAS application is installed for Windows x86, amd64 and x86_64.
24 April 2007	The project announces the campaign is coming to a successful end and that results are being prepared for the gallery and for publication.
May 2007	Science run ends; the Results Gallery is populated with 25 loaded tooltips, 25 unloaded tooltips and 3 method evaluation molecules.
2008	The project closes. Helfrich leaves Nanorex; Nanorex does not keep the project running.
12 December 2011	Allis announces the publication of the project's results on his website, thanking the volunteers.

Closure

No further campaign followed the tooltip survey. Helfrich's curriculum vitae records his tenure at Nanorex in Bloomfield Hills, Michigan as running from June 2006 to June 2008, during which he "created and managed a project to find failure modes for 25 published carbon-dimer mechanosynthesis tooltips" and "implemented a ~10,000 node, globally distributed computing network to run the project (nanohive-1.org/atHome)". A message posted to the BOINCstats forum on 2 October 2008 quoted Helfrich's reply to a participant's query about the project's future:

"Yes, sad to say, NHAH is dead. I no longer work at Nanorex and they have no interest in keeping it alive, so it's over."

Forum members reacted with regret, and asked whether the statistics could be archived, since the project had disappeared from the BOINCstats project menu into the retired projects section. As BOINCstats' operator explained in the same thread, retired projects continue to have user, team and country credit preserved, but host credit is no longer updated and retired projects are excluded from BOINC combined statistics.

Participation and statistics

NanoHive@Home was a modest project by credit, but a well-populated one by headcount for a company-run effort with a Windows-only client and a single science run.

BOINCstats project figures for Nano-Hive@Home

Metric	27 April 2007 (near peak)	14 January 2008 (after closure)
Users (total / active)	3,150 / 2,199	3,165 / 0
Hosts (total / active)	7,294 / 7,287	7,166 / 7,112
Teams (total / active)	452 / 368	454 / 0
Countries (total / active)	88 / 80	87 / 4
Total credit	41,865,304	42,983,633
Average performance	4,720.4 GigaFLOPS (4.720 TFLOPS)	0.0 GigaFLOPS

Other counts are cited by the participants themselves. The published paper's acknowledgement credits "over 6,000 worldwide volunteers and their computers", and Helfrich describes a network of roughly 10,000 nodes. Allis's curriculum vitae notes that the Tooltip Failure Mode Search Project "peaked at an unadvertised three teraFLOPS", while the contemporaneous BOINCstats snapshot records 4.72 TFLOPS. The discrepancy is typical of the period, when project-side FLOPS estimates and credit-derived estimates were computed differently. Sustained throughput can be estimated from granted credit using the BOINC convention that one credit corresponds to $10^{-2}$ cobblestone days of a reference machine, but the project itself simply published daily plots of results returned and average teraFLOPS per day on its front page.

Statistics were also carried by the usual third-party sites of the day, among them BOINCstats, Free-DC, BOINC.dk, boinc.mundayweb.com, DC-Vault, BOINC UK, the Knights Who Say 'Ni' and BOINC Combined Statistics.

Results

A conceptual rendering of a molecular nanotechnology device. NanoHive@Home's practical contribution was less glamorous and more useful: an exhaustive catalogue of the ways in which such structures can quietly break.

The Q-SMAKAS survey generated approximately 80,000 tooltip geometries which were used in 200,000 geometry optimizations at the RHF/3-21G or RHF/STO-3G levels. Those optimizations were catalogued, grouped by energy and geometry, and analysed to identify pathologies for both loaded and unloaded tooltips.

Several classes of defect emerged from the analysis:

• Hydrogen migration. In most stable tooltips, migration of hydrogen is assumed to be the most accessible route to an inoperative structure, a mode earlier work called "hydrogen poisoning". For the DCB65Ge tooltip, hydrogen insertion was the only pathology found, indicating an unusually deformation-resistant framework.
• Hydrogen inversion. The most common defect encountered was a C-H group on the outer surface of a cage flipping so that the hydrogen points into the hollow interior. The project frankly identified this as largely an artefact of the method, arising from the mismatch between light hydrogen momenta and the classical molecular dynamics time step, and described it with the memorable image of "reaching for your mouse with your right arm and instead firmly grasping your pancreas." Standard workarounds noted in the paper are smaller time steps, artificially increasing the hydrogen mass, or subsuming hydrogen into its heavy atom.
• Ge-C framework bond breaking, and the combined case of Ge-C breaking with formation of a C=C pi bond, which can produce very stable geometries and therefore demand large transition state barriers if the tooltip is to remain operable.
• Ge-Ge bond formation in unloaded tooltips, which is not necessarily harmful: by raising the barrier to other defects it may stabilize the tool during the interval between depositing one dimer and being recharged with the next.
• Conformational differences at the tooltip base. The paper singles this out as one of the more interesting outcomes, since controlling base conformation had not previously been considered as a route to tooltip design optimization.

One quantitative example given in the published discussion concerns the L1 to L2 rearrangement, in which the only identified loaded-tooltip pathology, a CH₂ torsion at the base opposite the dimer binding site, lies 1.3 kJ/mol below the operational geometry while the transition state for the rearrangement sits at 51 kJ/mol at the B3LYP/6-31G(d,p) level.

The method evaluation molecules were equally important to the project's argument. Running the same machinery at 300 K rather than 3000 K turned Q-SMAKAS into an ordinary conformational search, and the resulting surveys of biotin, n-heptane and n-octane demonstrated that the NanoHive@Home framework was useful for generating and optimizing molecular conformations in general, not only for torturing diamondoid cages.

The paper also drew a methodological lesson about volunteer computing itself, observing that both the speed and ultimately the quality of a calculation are dictated by the hardware the participants happen to own, and recommending that researchers survey the machines available to them before fixing the constraints of a quantum chemistry study.

Publications

The project's results were published four years after the science run ended, in the July 2011 issue of the Journal of Computational and Theoretical Nanoscience.

• (2011-07-01).Analysis of Diamondoid Mechanosynthesis Tooltip Pathologies Generated via a Distributed Computing Approach. Journal of Computational and Theoretical Nanoscience. pp. 1139-1161. DOI: 10.1166/jctn.2011.1792. Author affiliations: Department of Chemistry, Syracuse University (Allis); Helcorp, Maplewood, New Jersey (Helfrich); Institute for Molecular Manufacturing, Palo Alto, California (Freitas and Merkle). Indexed keywords include BOINC platform, conformational searches, diamondoid mechanosynthesis, distributed computing, molecular manufacturing, NanoHive@Home and tooltip pathologies.

Directly related papers

These are the source papers for the tooltip set that NanoHive@Home analysed, and the follow-on literature in which the project's results are cited.

• (May 2007).Horizontal Ge-Substituted Polymantane-Based C2 Dimer Placement Tooltip Motifs for Diamond Mechanosynthesis. Journal of Computational and Theoretical Nanoscience. pp. 433-442. The survey that supplied the 25 tooltips studied by the project.
• (March 2005).Design and Analysis of a Molecular Tool for Carbon Transfer in Mechanosynthesis. Journal of Computational and Theoretical Nanoscience. pp. 45-55. The DC10c tooltip study whose section 3.5 defect analysis Q-SMAKAS was designed to automate.
• (March 2004).A Simple Tool for Positional Diamond Mechanosynthesis, and its Method of Manufacture. Journal of Computational and Theoretical Nanoscience. pp. 71-80. Earlier work in the same tooltip programme.
• (2011).Tip-Based Nanofabrication: Fundamentals and Applications. Springer. ISBN 978-1-4419-9899-6. Provides background on the tooltip literature that NanoHive@Home extended.
• Freitas, Robert A..Nanomedicine and Nanorobotics Publications. rfreitas.com. Retrieved 2026-08-14. Lists the 2011 NanoHive@Home paper alongside the wider diamondoid mechanosynthesis literature.

Two documents from the original project website were re-posted as PDFs by Allis in 2011 for historical preservation: an overview of NanoHive-1 and NanoHive@Home, and the full Q-SMAKAS explanation.

People

• Brian Helfrich, project engineer and creator. Helfrich designed and implemented NanoHive-1 and created NanoHive@Home while Senior Architect and Software Engineer at Nanorex. He holds a BA in Computer Science from the University of California, Berkeley, began his career at Bellcore Labs, and later worked at Amazon, where he created the statistical model behind EC2 Spot Instance pricing. At Nanorex he also integrated an HDF5-based results library into GROMACS and led development of NanoVision-1.
• Dr. Damian G. Allis, project scientist. Allis took his doctorate in quantum chemistry at Syracuse University, received the Foresight Institute's Feynman Distinguished Student Award in 2004, served on Nanorex's advisory board, and worked on molecular tooltip development in coordination with K. Eric Drexler, Robert Freitas and Ralph Merkle. He wrote the project's public explanation of Q-SMAKAS and coordinated the scientific analysis. His CV dates his NanoHive@Home involvement to February to May 2007, with the project defunct in 2008.
• Robert A. Freitas Jr. and Ralph C. Merkle of the Institute for Molecular Manufacturing, co-authors of the resulting paper and authors of the underlying tooltip survey.

Legacy

NanoHive@Home is one of relatively few BOINC projects to have carried an entire published quantum chemistry study from work unit to journal page, and it is unusual among them in having been organized by a private company that then dissolved the effort. Its lasting contributions are the peer-reviewed pathology catalogue, the demonstration that a molecular dynamics plus quantum chemistry pipeline can be split across tens of thousands of consumer machines, and the observation that the same protocol run cold serves as a general conformational search engine.

Its practical successor in subject matter is nanoHUB@Home, a later BOINC project supporting nanoscience and nanotechnology research at nanoHUB.org, though the two share no code or personnel. Chemistry-oriented BOINC projects that followed similar quantum chemical or conformational search strategies include QMC@Home, QuChemPedIA@home, Leiden Classical, eOn and Spinhenge@home. NanoHive-1 itself remains available on SourceForge.

The project also survives as a small piece of volunteer computing folklore. Its screensaver appears in period discussions as a benchmark for how engaging a BOINC display could be, and threads announcing its beta rounds on AnandTech and BOINCstats preserve a detailed record of what it felt like to test a young BOINC project in 2006.

See also

• BOINC projects
• BOINC project screensavers
• nanoHUB@Home
• QMC@Home
• QuChemPedIA@home
• Spinhenge@home
• Leiden Classical
• eOn
• Predictor@home
• SETI@home
• Rosetta@home

External links

• NanoHive@Home home page (archived, 5 May 2007) via Wayback Machine
• The Q-SMAKAS Tooltip Failure Mode Search Project, by Damian G. Allis (archived) via Wayback Machine
• NanoHive@Home Results Gallery (archived, 25 August 2007) via Wayback Machine
• Simulation Software and BOINC architecture (archived) via Wayback Machine
• Screensaver Graphics explained (archived) via Wayback Machine
• NanoHive Nanospace Simulator on SourceForge
• NanoHive@Home's Published Results, somewhereville.com, 12 December 2011
• Publication record at IngentaConnect
• BOINCstats forum thread on Nano-Hive@Home (retired projects)