Opening Doors at the Atomic Level

45 nm technology is already seeing the addition of new materials to the process line, with the thinnest layers approaching a thickness of 3–5 atoms in high-performance devices. An atom (~2 Å in diameter) out of place can potentially cause device failure. Therefore, along with materials and integration challenges, characterization techniques must be able to locate and identify single atoms, a task akin to searching for a needle in a specific car somewhere in the United States.

Working to achieve atomic-level understanding and control of the electronic properties of interfaces means exploring new materials and finding the diagnostics to better manipulate electron and spin transport, said David Muller, a professor in the department of applied and engineering physics at Cornell University (Ithaca, N.Y.) and team leader at the Cornell Center for Material Research (CCMR).

The CCMR supports four interdisciplinary research groups (IRG) and a number of smaller "seed" research groups through a National Science Foundation (NSF) MRSEC grant and university support. With faculty members spanning 12 departments and four colleges, and collaborations among academic and industrial researchers worldwide, work is being carried out that can impact the semiconductor industry 10 years out.

Three projects of particular interest focus on rare earth materials, generating some unexpected results; the development of molecular electronic devices; and improvements in electron microscopy that may provide the insight needed to realize devices several technology nodes into the future.

Electron microscopy

The state-of-the-art method for imaging at the atomic level is a 50-year-old technology: electron microscopy. Because of the nature of electric and magnetic fields in a vacuum, the cylindrically symmetric lenses designed for electrons have severe spherical aberrations. Wavelength for wavelength, the standard lens system has essentially the same optical properties as a beer bottle, forcing scientists to use tiny numerical apertures (NA) of ~0.01 to obtain distortion-free images. In theory, the spatial resolution of electron microscopes can be improved by a factor of 50 with a lens system that can eliminate these inherent spherical aberrations.

Now, after 50 years, a revolution is underway that provides an optics solution. The new generation of electron microscopes has a factor of 2–4 higher resolution and a factor of 10–100 more useable current, making them 10–100× faster. This was achieved with corrective optics that consists of more than 80 multiple pole lenses placed in the cylindrical lens system.

Faster, higher quality imaging has made it easier for the CCMR researchers to study device failures and distinguish between intrinsic limits and improper crystal growth. The investigation of circuit degradation, such as dopant diffusion through gate oxides, unintentional channel doping, barrier breakthrough, and line edge roughness, has also been studied in collaboration with students supported by the Semiconductor Research Corp. (Durham, N.C.). To image these regions, a thin slice is carved from a device with a focused ion beam (FIB). In Figure 1, a Ta/TaN liner that will be filled with copper is clearly viewed. Similarly, in Figure 2, roughness and thickness variations in the sidewalls of four copper-filled trenches is shown in 3-D.

1. In the electron microscopy image, the gold color represents the outer surface of a Ta/TaN liner. The plane containing the red line is a cut through the entire via, showing the tantalum and TaN layers.

When concentrations exceed 10¹⁹ at/cm³, the probability of finding a dopant atom within a few atoms from another becomes very high, allowing them to combine, share available electrons, and electrically deactivate. These "dopant clusters" can potentially become problematic as dopant densities increase to maintain low resistivities in small geometries. Their impact on device failure is being studied with electron microscopy.

2. The 3-D electron micro­- scopy image of copper-filled trenches indicates roughness in the sidewalls. The red outlines a density slice through the line, useful for measuring sidewall thickness.

Electron microscopy has always been a powerful tool. The optical improvements make it a diagnostic technique for use on a daily basis for failure analysis, as opposed to per experiment.

2-D interfacial layers

Looking ahead, perhaps three or four technology nodes to below 10 nm will require the development of new materials and device design strategies to a far greater extent than what is taking place today. At these geometries, a new class of phenomena occurs, which can be thought of as interface phases. These structures and properties of materials do not exist in bulk material. However, when some materials that are insulators in bulk combine with similar materials, a stable 2-D conducting layer can be formed at the interface.

CCRM researchers have helped colleagues, such as Prof. Harold Hwang at the University of Tokyo and Prof. Jochen Mannhart at the University of Augsburg in Germany, to characterize highly mobile electron systems at the interface of insulating perovskites oxides, rare earth titanates, and aluminates such as strontium titanate (SrTiO₃), lanthanum titanate (LaTiO₃) and lanthanum aluminate (LaAlO₃).

Samples were prepared by depositing a thin layer of SrTiO₃ (between 5 to 100 atoms thick) on an equally thin layer of LaAlO₃, grown with pulsed laser deposition at 770°C in ultrahigh vacuum. Contacts were formed using standard lithographic processes and the structure biased. The result was unexpected. At the interface, a couple of atoms deep, the researchers measured a very high density of conduction electrons, a factor of 20 higher than the most highly doped silicon device, Muller noted. Yet by switching one layer of atoms at the interface, it became insulating.

3. A high-angle annular darkfield image of a LaAlO3 film grown on SrTiO3 shows a coherent interface.1

The results have created much discussion as to the origin of the conducting layer: Is it due to the oxygen vacancies in the SrTiO₃ crystal, or is it related to the polar nature of the LaAlO₃ structure wherein a potential develops during the growth of the layer leading to an electronic reconstruction above some critical thickness?²

While insulating perovskites oxides are not intended to replace silicon, they do embody characteristics not found in silicon (in particular, magnetism and ferroelectricity), perhaps lending itself to all-oxide devices such as ferroelectric memories or spintronic devices.

Growing crystals with atomic-level control and the exploration of novel materials is only the beginning. "We are at the earliest of stages of making something useful," Muller said.

Molecular electronics

"The sensitivity of electrons to their local environment provides the basis for building electronic functionality into molecular architectures," said Héctor D. Abruña, a professor of chemistry and an IRG co-leader, along with Dan Ralph, a professor of physics. To take advantage of this sensitivity, the group's focus is on understanding the transport behavior of matter at the nanoscale by trafficking electrons and photons at interfaces. This is done by placing a single molecule between two contacts.

Break junctions are used as the electrodes for contacting the molecule (Fig. 4). A tiny metal (often gold) bowtie structure is formed using nanofabrication techniques. Residing between the two contacts is a narrowed region where a monolayer of some molecule is absorbed. This occurs when current flows through the bowtie, locally heating and melting the middle region and forming a gap. With a bit of luck, a single molecule falls into the gap.

4. This schematic illustrates the formation of a gap used to capture a single molecule in a break junction-based single molecule transistor.

Mechanical break junctions resemble a drawbridge where, depending on the amount of pressure, the gap between the two electrodes opens and closes, thereby changing the coupling of whatever resides in the middle. In this case, it is a C₆₀ molecule. The CCMR researchers were able to wire single molecules between the two contacts using field-effect transistor (FET) geometry. By having access to a gate voltage, they could easily distinguish between a good molecular contact and a bad short. Moreover, the IV curve is dependent on gate voltage, providing an unmistakable signature of a molecular event, Abruña noted. It makes the experiment more difficult, however, as there is no question when a molecule is in the gap.

Simple organic molecules, such as the benzene rings that are coupled with triple bonds, are typically used in these experiments. The problem is that the addition or removal of electrons from those structures makes them unstable. The molecules also require bias voltages on the order of 5 V that, when dropped over a few angstroms, generates an enormous field.

To overcome the limitations of organic molecules, the CCMR group uses transition metal complexes, which are molecules that have metal centers and are stable in multiple oxidation states. This means that the molecules are stable and electron counts can be controlled, making it possible to go from species that are paramagnetic or diamagnetic to allow the study of magnetic field effects.

Although progress is being made in the field of molecular electronics, there are some inherent problems that may limit its range of application. For example, how do you get amplification in these devices? Instead, sensor applications stand to gain greatly from these advances. Perhaps through the perceptive use of molecular assemblies, functionality can be added in surprisingly unique ways.