Hidden imaging data Magnetic resonance imaging (MRI), which is based on nuclear magnetic resonance (NMR), has proved to be one of the more pervasive applications of physics to medical technology. MRI can probe softtissue characteristics that X-rays and ultrasound cannot detect because it measures chemical conditions in tissues. But improving MRI resolution, now around a few millimeters, poses a major challenge. Higher resolutions require higher magnetic field gradients, and gradients much higher than those now used cause side effects on patients’ nervous systems. But a collaboration between Valerij G. Kiselev of the section of medical physics at the University Hospital Freiburg (Germany) and Dmitry S. Novikov of the physics department at the Massachusetts Institute of Technology has shown that information down to the micrometer scale lies hidden in NMR data that is generally thrown away. Although their new technique does not directly increase resolution, it gives information about the shapes of micrometersized cells and capillaries in a given millimeter- sized region. Such shape information can be vital for the early detection of several conditions, including tumors.

Resolution of magnetic resonance images, currently around a few millimeters, may be improved down to the micrometer scale by retrieving shape information from signal decay rates. (Tom Moore)

In conventional NMR imaging, a strong magnetic-field gradient allows only a single millimeter-sized pixel at a time to be identified by a pulse of radio-frequency (RF) radiation. The radiation flips the magnetic spins of hydrogen nuclei aligned by the magnet field. The effect is akin to a hammer hitting a tuning fork, and the nuclei emit a rapidly fading RF signal as they precess around the field direction. The rate at which the signal fades depends on the physical and chemical environment of the nuclei. For example, the fading time for tumors differs from that for healthy tissue.

“The conventional approach assumes that the fading is a steady exponential decay, and measures the rate after enough time so that is nearly true,” explains Novikov. “But our calculations showed that for relatively short times after the pulse, the decay rate is slower, and the way the decay rate changes contains valuable information.”

Specifically, the researchers looked at how this early time-fading changes depending on the microscopic shape of the objects that contain the nuclei. For fat objects such as spheres, the decay rate increases rapidly. For thin objects such as rods or thin disks, it increases more slowly. These changes stem from the different ways that microscopic objects alter the magnetic fields that affect the nuclei. “A skinny object produces a more uniform field, while a fat object’s field is less uniform,” says Novikov.

Experiments and simulations comparing the signals from microscopic polystyrene spheres and those from disk-shaped red blood cells have closely agreed with the team’s calculations. Potentially, the ability to distinguish, for example, the average shape of capillaries in the brain would have great diagnostic value. When a tumor begins to form, normally straight capillaries become curled, a shape change that, in theory, the new method would detect.

However, much work remains before physicians can apply the new analysis in practice. “It’s much easier to predict the decay-rate pattern knowing the shape than to determine the shape from the decay-rate pattern,” explains Novikov. More experiments will be needed to look at the differences in the evolution of decay rates in tumors and normal tissue, as well as more detailed calculations of how decay rates vary with different tissue conditions.

To create three-dimensional integrated circuits, it is possible to precision-bond two wafers by first bonding one wafer to glass and removing the silicon substrate to make it transparent, as in this 200-mm stack. (Z. Siwy and A. Fulinski)

Piggy-back chips All computer chips today function as essentially two-dimensional arrays, with devices placed next to but not atop one another. Such an arrangement has a significant disadvantage. The number of devices on a chip grows as the inverse square of the size of the devices, but the number of connections to the chip can grow, at most, linearly with the inverse of the size. As a result, more and more devices compete for the same communication lines on and off the chip, and communication delays become more important than processing delays.

One solution to this problem is to manufacture three-dimensional integrated circuits (3D ICs) with devices stacked in piles. Connections between one layer and the one underneath can now grow as rapidly as the number of devices and be very short as well, significantly increasing overall speed compared with two chips side by side. Manufacturing such 3D ICs is no easy task using conventional lithographic techniques, which add one layer of material at a time to a substrate. For one thing, piling up layers introduces more and more dislocations. For another, some processes needed for one layer will damage a completed device in the layer beneath it.

An IBM research team led by Kathryn Guarini, a research staff member at the Thomas J. Watson Research Center (Yorktown Heights, NY), has now developed a way that may make 3D ICs practical. It consists of producing two conventional wafers separately and then bonding them one on top of the other. Both wafers are produced conventionally, and then one is bonded with polymer adhesives through its top plane to a glass wafer. The original silicon substrate of this wafer is then carefully removed by etching and mechanical grinding, stopping at the silicon oxide insulating layer. The combined wafer and glass plate is now transparent, so it can be exactly aligned with the other chip wafer and bonded to it. Then laser ablation releases the glass layer and removes the adhesives. In a final step, which the team has not yet performed, holes can be etched through the two wafers and filled with metal to produce the large number of connections that will make each pair of chips function as one.

“What we have demonstrated is that this method is gentle enough to preserve the high performance of the devices,” Guarini says. Performance increases of around 30% are possible using just two layers, with further increases accruing as subsequent layers are added.

Scanning electron micrograph of the 500-nm-wide end of a conical pore formed in 12-µm-thick plastic film, part of a synthetic ion pump that moves cations against a concentration gradient.

Synthetic ion pump The process of separating chemicals and concentrating desired species on one side of a barrier is fundamental to industry and life itself. In industry, separation generally involves physical processes such as evaporation, chemical reactions that select specific ions or molecules, or reverse osmosis using membranes that allow passage of solvents such as water but not their dissolved substances. Cells, however, rely mainly on ion pumps, which consist of tiny pores in cell membranes that selectively move ions from less concentrated to more concentrated regions.

Zuzanna Siwy of the Gesellschaft für Schwerionenforschung (Darmstadt, Germany) and Andrzej Fulinski of the M. Smoluchowski Institute of Physics at Jagellonian University (Krakow, Poland) have now developed a synthetic ion pump that is based on tiny conical pores formed in a 12-µm-thick plastic film (Phys. Rev. Lett. 2002, 89, 198103-1). The pores act as miniature ratchets, and an oscillating electrical field preferentially pushes cations (positively charged ions) from the narrow to the wide ends of the cones.

Siwy and Fulinski were studying biological channels and thought that synthetic pores would provide a way to model the biological phenomenon. They got the idea that external field oscillations, which had been observed in the biological pumps, could be used to create synthetic ion pumps. “We got the hint that a conical pore could work as a ratchet after we saw an article in Scientific American about molecular motors,” Siwy explains. Molecular-sized ratchets work on the principle that a random or oscillating impulse can be converted into useful work if a device can move only in one direction.

A conical pore works as a ratchet because the pore, created in poly(ethylene terephthalate) film, is lined with negative charges. These charges are created during the etching of the pores, which breaks the polymer chains and leaves the negative charges exposed. A positive ion is attracted to the concentration of negative charge, which is highest at the narrow end of the pore. In the absence of an additional field, no flow would result because ions would just cluster around the pore’s narrow end until the charge was neutralized.

However, an external field imposes a 0.5-V oscillating potential on the membrane. Positive ions fall into the deep potential well near the narrow end, which draws them across the membrane into the pore. The potential well is asymmetrical, sloping steeply toward the narrow end and gently toward the wide end of the pore. Inside the pore, the field created by the oscillating potential occasionally pushes the ions with enough force to move them out of the well on the gentle-sloped wide end, but never enough to overcome the steep narrow end. Thus, the ions are prevented from recrossing the membrane and must exit out the wide end. A net flow of ionic current of up to 3 billion ions/s flows through the 2-nm-diameter narrow end of the pore and out the 500-nm-diameter wide end.

The efficiency of the device is high—40% when equal amounts of ions are on both sides of the membrane—and it falls to 10% for a 7.5-fold difference in concentrations. Some flow is observed even when the concentration of ions on the receiving side is 100 times higher than on the contributing side.

The conical pores are formed by irradiation through a 100-mm-diameter mask with a 2.2-GeV beam of gold ions from the UNILAC linear accelerator in Darmstadt. The irradiation is stopped as soon as a single ion passes through the membrane. The irradiated material is then etched away, and the etching is stopped as soon as a breakthrough of the membrane occurs, producing the atom-sized narrow pore.

“We are now working to develop a device that can enrich solutions on a significant scale, using many pores,” says Siwy. This effort requires a number of improvements: making the pump work faster, designing pores that work for negative ions as well as positive ions, and determining if the pores’ efficiency varies for different ions.

Transmission electron micrograph of a new form of copper that is six times as strong and nearly as ductile as conventional copper and is formed by modifying its microstructure. ( Johns Hopkins University)

Strong, ductile copper Copper’s usefulness is based on its high electrical and thermal conductivity, and its high ductility: the conventional form of copper can undergo a strain of almost 70% before breaking. Such ductility makes copper easy to form and means that it yields gracefully rather than catastrophically to mechanical stress. But copper is not strong, with a strength of about 600 kg/cm2 before starting to deform. This relatively low strength is a drawback for the growing field of microelectromechanical system (MEMS) devices, which require high strength to compensate for their tiny size. As with many materials, copper’s mechanical properties are in part determined by its microstructure, which produces characteristic defects that strengthen or weaken the metal. Conventional annealed copper has relatively large grains several micrometers across. A team of physicists at Johns Hopkins University has invented a method of modifying its microstructure to produce a new form of copper that is six times as strong and nearly as ductile as the conventional metal (Nature 2002, 419, 912). The key is to create a mixture of extremely small and larger grains.

The research team reasoned that higher strength could be obtained with nanocrystals smaller than 200 nm in size, so they worked copper at room temperature without heating it to reduce crystal size. (Such cold working was the key to the strength of the fabled Arabian steel swords of the Middle Ages.) However, the team found that although strength increased greatly, ductility practically disappeared, with almost no uniform deformation and only 10% strain before failure. “We realized that the dislocations that increase strength were being eliminated by recrystallization,” explains Mingwei Chen, a member of the team. “So we decided to try cold working at liquid-nitrogen temperature to prevent recrystallization.” This approach increased strength to almost 5,000 kg/cm² and allowed a small uniform deformation, but ductility was still minimal.

The problem was that the large grains, although relatively weak, allowed the ductile deformation that strong but brittle nanocrystals could not. So the researchers heated the low-temperature-worked samples to 200 °C, producing the desirable mix of strong nanocrystals and ductile large crystals. At this point, strength was still 4,000 kg/cm² but uniform deformation extended to almost 30% strain, and failure occurred only at more than 65% strain, almost as much as for conventional low-strength copper.

Not only could such strong and ductile copper prove useful for MEMS devices, it could have broader applications where mechanical strength and high conductivity are required. In addition, the researchers believe that the same combination of low-temperature working and high-temperature annealing can improve the toughness of many other nanocrystalline materials.