Hydrogen@Home is a research project that uses Internet-connected computers to do research in Hydrogen Production. Our project is in a conceptual developement phase called "Alpha" Phase, you can participate by downloading and running a free program on your computer. Visit the project here - http://hydrogenathome.org/

Have you seen all the media attention paid to hydrogen fuel cell vehicles? Instead of emitting Greenhouse Gases such as CO2, Hydrogen Fuel Cell Vehicles release water vapor. Tank-to-wheel efficiency of a fuel cell vehicle approaches 45%, compared to diesel vehicles which are 22%.

Fuel cell technologies work by reacting chemicals in a manner that forces electrons to travel through a circuit. Hydrogen fuel cells work by reacting Hydrogen with Oxygen which results in water vapor and energy. Many people believe fuels cells are the solution to Global Climate Change, given the abundance of hydrogen in the form of water.

Extracting hydrogen from water is an expensive process using electrolysis, the process of passing electrical current in water. This just export the problem to the power plant. Both fossil fuels and nuclear power have their respective problems.

We could go through a litany describing all potential renewable energy technologies. Among the major categories are solar, wind and geothermal. Despite years of study, none of these technologies have managed to compete on the same scale as fossil fuels or nuclear power.

Biological systems have the singular ability to self-optimize, replenish and require minimal resources. Provide a pool of algae with water, light and a good balance of nutrients and they will thrive in temperate climates. In fact, there are even reports of algae thriving under Arctic Sea Ice! So many of our medicines are manufactured through biological systems: antibiotics and vaccines. Devices used to grow these medicines and other biologically fabricated materials, are frequently called bioreactors.

In the late 1990s, UC Berkeley scientist Anastasios Melis experimented on algae cultures and discovered sulfur deprived algae can produce hydrogen through photosynthesis. Dr Melis discovered the enzyme hydrogenase is responsible for catalyzing this reaction and that this enzyme looses this function in the presence of oxygen. Algae producing this hydrogen are exhausted as oxygen quickly accumulates and causes damage to the hydrogenase.

Any technology analysis of a Hydrogen economy must address the following issues: Production, Application, Storage, Transport and Assess the environmental impact (PASTA). Since the US Department of Energy (DOE) set the Corporate Average Fuel Economy using an acronym called CAFE, it seems fitting that we can reduce this analysis to an acronym called PASTA.

In 2004, the DOE published the "Updated Cost Analysis of Photo biological Hydrogen Production from Chlamydomonas reinhardtii Green Algae", to project the economic viability of biological hydrogen production. Their 2004 analysis estimated it would cost $13.53 USD per kilogram which is roughly the energy equivalent to a gallon or 4 liters of petrol.

Recent developments suggest that the above cost analysis is less relevant because it does not take into account improved efficiency of light conversion through genetic mutations. In 2007, Dr Melis reported "~15% utilization efficiency of absorbed light energy was achieved", by mutating the genes responsible for Chlorophyll Antennae structures. Naturally occuring algae only achieve 0.2% efficiency.

Despite all this good news the question remains, how can we mitigate oxidative damage to Hydrogenase? Is it possible to computationally model biological hydrogen production and use these models to mitigate this oxidation? Hydrogen@Home is developing such models.

In the short term Hydrogen@Home is interested in computational research on biological hydrogen production, particularly how this could scale economically. Hydrogen@Home is actively pursing collaborative relationships with several academic institutions, government agencies and corporations.

Our broad vision for Hydrogen@Home is to analyze aspects of renewable hydrogen technology described by PASTA. Computationally modeling hydrogen production is a difficult task requiring many theoretical considerations. In the short term, we intend to build accurate models of these biochemical interactions. Long term, we seek an experimental design that will screen chemical data for new catalysts (organic or inorganic) useful in hydrogen production. As a way to learn about BOINC and computational chemistry, several open source molecular modeling tools have been employed.

In order to model thermodynamic free energy characterized by substrate binding to an enzyme's active site, Hydrogen@Home employs Molecular Docking simulations. Molecular Docking is a type of simulation best suited for identifying potential drugs for diseases as demonstrated by FightAids@Home where researchers try to find molecules that disrupt the normal function of HIV Protease. Mark Somers developed a molecular simulation user interface on Leiden Classical, which inspired me to implement a user interface that facilitates custom docking simulations. This way users can submit their own Docking Experiment and possibly replicate experiments performed by FightAids@Home. Potentially, such interfaces could end up in science classrooms.

Substrate binding to enzymes is only part of the hydrogen modeling odyssey. Molecular Dynamics is a modeling approach where a computer attempts to model molecular movements or atomistic interactions over a length of time. As we have learned more about molecular modeling, we understand that Hybrid QM/MM Molecular Dynamics simulations provide the best potential for modeling enzyme interactions on a reasonable timescale. Hybrid QM/MM applies computationally expensive quantum theory over a limited area and computationally less expensive classical physics to a larger area of the simulation. Classical physics alone, is not effective in modeling these enzyme catalyzed chemical reactions, applying quantum theory to a limited area makes it computationally feasible.

Are these simulations feasible for distributed computing? Before we can effectively design these Molecular Dynamics simulations, we must determine the best experimental methods; What computational methods are reasonable and accurately predict outcomes that match reality. We should also define how small these molecular dynamics calculations can be reduced for distribution over the Internet.

In general, CPU and memory requirements are:

http://anusf.anu.edu.au/%7Evvv900/qm-mm.html

There are four theoretical assumptions that we can apply independently to any molecular dynamics simulations: ab initio, Density Functional Theory (DFT), Semi-Empirical and Classical Molecular Mechanics. We could design ab initio simulations that applies pure quantum theory and is computationally expensive. The computational cost of DFT can vary depending on theoretical considerations, they model the complete electronic structure of molecules. Semi-empirical QM methods are computationally more efficient on orders of magnitude than DFT or ab initio methods because they use some predefine parameters called Molecular Force Fields; however, these force fields can be difficult to derive and do not always transfer from one experiment to another. Finally Molecular Mechanics is the simplest method of modeling molecular interactions and has many applications.

Example: One Energy Evaluation for a Peptide with 126 Atoms:

*Semi-empirical PM3 method http://anusf.anu.edu.au/%7Evvv900/qm-mm.html

If you consider the above semi-empirical an N² and ab initio N⁴, an equivalent ab initio would take approximately 1 day and 23 GB of memory. Perhaps with a multi core processor, these calculations can be achieved quicker.

As we go forward developing our scientific method, we should study these molecular force fields applied to Semi-Empirical QM. Molecular Force Fields are essentially derived from experimental data related to harmonic force constants for molecular bonds. Some of this data is available to the public domain in either standardized formats or in journal articles, while certain chemical companies maintain proprietary databases of this empirical data. There may be some ab initio methods of calculating these parameters but it is unclear to this author how computationally expensive this will be given the proprietary nature of many chemical software packages.

At least one open source package exists that has a potential benefit in this analysis. JavaGenes is a java based genetic algorithm developed by NASA researcher Albert Globus. In his paper published by NASA Ames Research Center, Globus outlined a method of automating the parameterization of a molecular force field and distributing this computational load: "We hypothesize that this step can be automated by large computations on cycle-harvested desktop computers. By automating parameterization, exploration of functional forms should be enhanced."

Assuming we can predict the functional forms, what would these molecular dynamics simulations look like? Molecular dynamics can be used to interpret empirical data or predict outcomes. These simulations provide data that characterizes free energy change and coordinates. We would like to determine the reaction mechanisms for hydrogen production or the Minimum Energy Path (MEP) for a reaction. There are several theoretical approaches to predicting MEP. One of the most frequently cited methods is called Nudge Elastic Band (NEB) method, which interpolates the energy changes by identifying reaction paths that provide smooth energy curves.

Once we have an accurate computer model that corresponds to reality, we can begin introducing variables in computer simulations and compare these predictions to experiments. If this methodology proves effective, we scale these experiments to a new level and explore many variables such as different enzymes, co-enzymes and alternative reaction pathways. If these methods are effective in modeling hydrogen production, we can reasonably assume they would be effective in modeling all aspects of PASTA. There is reason to believe fuel cell technology can achieve 90% tank-to-wheel efficiency using enzymes that facilitate proton transport. In this scenario, vehicles could travel twice as far on one kg of Hydrogen.

Besides modeling the actual enzyme catalyzed reaction, another less computationally demanding approach is identifying proteins with structural homology to Hydrogenase and vectoring their genes into the algae in a way that they are delivered to the same reaction site. This should be a relative easy approach and it could be implemented as one group of workunits run on Hydrogen@Home.

There is much work to be done before we can achieve a Hydrogen Economy. We do believe this research will eventually succeed with or without these computational models. With a good experimental design, there is a chance we can get there sooner!

Visit the project here - http://hydrogenathome.org/