Manipulating Atoms
Manipulating AtomsShown above: Valence charge density distributions shown for two atomic layers of iron on a magnesium oxide (MgO) substrate. Maximum distribution is represented by red; midrange is yellow. Minimum charge density is in darker blues. (Courtesy of Arthur J. Freeman and Chun Li)
Medieval alchemists dreamed of transmuting common base metals into gold--such miraculous conversions proved impossible, of course. Today, thanks to discoveries in the fields of computational physics and computational chemistry, researchers are making startling progress in understanding and manipulating matter at the atomic and molecular level.
While no one yet has figured out how to turn lead into gold, advances made possible by high-performance computing (HPC) have enabled scientists to use the elements provided by nature to create entirely new artificial materials with properties tuned to meet a variety of mankind's needs. A premier example can be seen in the work of Arthur J. Freeman, NCSA principal investigator and Morrison Professor of Physics at Northwestern University.
He is also doing additional research on other types of materials, such as aluminum intermetallics used in the aerospace industry.
The Northwestern professor is quick to point out that the opportunities for computational physics research are ``as broad as they are deep. We deal with all kinds of materials, without restriction. Ultimately, we hope that new understanding of all of these materials will be transferred to making new materials with properties that [mankind] can use.''
Due to the enormity and complexity of the calculations involved, quantum theory defies direct solution. For example, a direct calculation to determine electron orbitals in a solid would require solving a large set of differential equations for each orbital of around 10-23 electrons in the material. As a result, various methods based on approximations are required that can reduce the number of equations, but they can also throw doubt on the results.
Today's supercomputers, capable of billions of calculations per second, have opened a new frontier. According to Freeman, computational physicists can now solve electronic structure theory problems to much higher degrees of precision, thanks to these machines. They are also able to tackle problems 100 to 1,000 times more complex than those previously considered.
``The ability to do calculations and simulations rapidly has meant, for us, a revolution,'' he relates. ``Over the years, methods such as the density functional theory of Hohenberg, Kohn, and Sham [1964-65] have evolved and developed. It's only because of access to supercomputers that we have been able to do calculations without approximations, other than the local density approximation inherent in current applications of the method.''
Based on a local density functional theory approach, the FLAPW method provides a highly precise method for calculating the electronic structure and properties of materials. FLAPW's key feature is its ability to solve the Poisson equation in 3D. (The Poisson equation relates the electrostatic potential of a material to its electron distribution.) Despite the fact that the potential acting on electrons is 3D, previous solutions were limited to a 1D potential to make the calculations manageable.
Freeman and associates have invested about 18 person-years in the development of FLAPW codes for Cray supercomputers. Initial work with the code was done at Lawrence Livermore National Laboratory; enhanced versions are used at NCSA and elsewhere by Freeman and a growing research community.
Using FLAPW, Freeman and associates have made some innovative discoveries. In the early 1980s, they predicted that the level of magnetism seen in common magnetic materials would be much greater at the surface than in the bulk of the material. In iron, for instance, the so-called magnetic moment of surface layer atoms was predicted to be 2.98 Bohr magnetons--about 40% higher than that measured in bulk atoms. In chromium, the difference is even more dramatic: at 2.5 Bohr magnetons at the surface versus 0.6 Bohr magnetons in the bulk. These predictions were confirmed by experiments at the University of California at Berkeley.
Through FLAPW simulations, the Northwestern researchers discovered that magnetism can be even further enhanced when materials are formed into single atomic layers. They found that the deposition of such monolayers onto certain other materials does not significantly decrease the effect. A monolayer of chromium deposited on gold, for example, produces a magnetic moment of 3.7 Bohr magnetons. ``So now you're up to six times larger than what the chromium moment is in the bulk,'' Freeman says. ``That's truly a giant moment.''
Recent predictions by Freeman and others show that elements that are not magnetic in their natural state can be made to exhibit magnetism when deposited in monolayer form on a substrate.
``Vanadium, which is just to the left of chromium in the periodic table, is not magnetic--either in bulk or at the surface. But if you put a layer of vanadium onto silver, it becomes magnetic with a sizeable magnetic moment. So here, we've made magnetism in a material that's never been magnetic before. For the other materials, we've enhanced the magnetism dramatically--40% for iron and 300% for chromium on the surface. For chromium on gold, the increase is 500%.''
Though still theoretical, these discoveries take on increasing importance given the wider availability of advanced equipment such as molecular beam epitaxy (MBE) systems that can be used to synthesize actual physical samples of monolayers (on benign substrates) that were first simulated on a supercomputer. ``These beautiful machines represent a new frontier which is driving the field of artificial materials,'' Freeman notes.
Beyond magnetic properties, the electrical characteristics of monolayers and ultrathin films are also of great import, particularly in the semiconductor industry. In other work with the FLAPW code, Freeman is simulating mechanisms found at the so-called heterojunctions formed in advanced electronic circuits at the interface of two materials. He is also studying the unusual properties of superlattices--sandwich-like materials made by alternating layers of metallic, semiconducting and/or insulating elements.
Another ``hot'' field is that of high-temperature superconductors. Early in 1987, Freeman and his collaborators were among the first to use supercomputers to investigate the properties of a newly discovered class of these materials consisting of complex yttrium- barium-copper-oxide compounds. These new superconductors (so-named because of their ability to conduct electricity with no resistance, while generating strong magnetic fields) operate at temperatures much higher than previously thought possible. As such, they hold potential for revolutionary advances in computers, electric motors, and magnetically levitated trains.
When copper-oxide materials were first discovered, researchers around the world proposed a variety of complex theories to explain high-temperature superconductivity. Through FLAPW simulations of the new materials, Freeman and his group were able to make predictions which were later proved out in laboratory experiments by others.
``I'm happy to say that of all the theories that were proposed about the normal state, the only theory that still survives is electronic structure theory,'' Freeman says. ``Thanks to supercomputers, we were able to respond quickly to the 1986 Nobel Prize-winning discoveries of Bednorz and Mller and later Chu and collaborators [of high-temperature superconductors] and to investigate the most complex materials we had ever come across. We showed how the crystallographic structure, the chemical and physical properties, and the electronic structure are related.''
Given the complexity of his research, it is not surprising that Freeman is one of NCSA's major NSF-supported users. He is also a major user at the Pittsburgh Supercomputing Center. The Northwestern group is constantly working to make their codes run more efficiently on the Crays and is working on enhancements which will ``allow us to do a better job with many-body problems.'' They plan to ``parallelize'' FLAPW for use on the Thinking Machines' CM- 5, NCSA's massively parallel processing machine.
On bestowing the award, Russian academician S. A. Vonsovsky cited the ``intensity'' of Freeman's work in ``hot'' fields and utilization of ``novel computational techniques.'' In 1992 Freeman was elected to the Academy of Natural Sciences of Russia and the Russian Academy of Sciences (formerly the Soviet Academy of Sciences).
``When people ask me why I'm so enthusiastic and why I still work so hard, it's because of the opportunity to do what we couldn't dream about, even a few years ago--the advances having been so enormous and so exciting,'' Freeman says. ``The sky is really the limit.''
__________________________________________________________________
access / Spring 1993 / NCSA / pubs@ncsa.uiuc.edu