LBNL Masthead A-Z IndexBerkeley Lab mastheadU.S. Department of Energy logoPhone BookJobsSearch
Looking at Transistor Gate Oxide Formation in Real Time Print


The oxide gate layer is critical to every transistor, and present-day layer thicknesses are in the 10–20 Å range (1–2 nm). However, little information exists on the oxidation process at this thickness. Available results are either for thicker layers grown under high-pressure conditions or for only the first couple of monolayers studied under high-vacuum conditions. Now, for the first time, a group of researchers has obtained real-time oxidation results for this elusive range. Using the ambient-pressure x-ray photoelectron spectroscopy (APXPS) endstation at ALS Beamline 9.3.2, they examined oxidation of Si(100) at pressures up to 1 torr and temperatures up to 450 ºC. The Si 2p chemical shifts allowed determination of oxide thickness as a function of time with a precision of 1–2 Å. The initial oxidation rate was very high (up to ~234 Å/h). Then, after an initial oxide thickness of 6–22 Å was formed, the rate decreased markedly (~1.5–4.0Å/h). Neither rate regime can be explained by the standard Deal-Grove (D-G) model for Si oxidation. These results are a significant step toward developing a better understanding of this critical thickness regime.

Top: Gate oxide layer atop layer of silicon oxide, resting on silcon substrate. Bottom: Atomic model of the Si–Si oxide interface, with the various chemical species indicated.

Smaller, Faster, Thinner . . . Better

Every day, approximately a hundred trillion individual transistors (about 4,000 per person) are produced worldwide as part of the integrated circuits that drive information technology. Each of these transistors contains an ever-thinner gate oxide layer, a crucial element in the ubiquitous field-effect transistor. It sits between the gate electrode and the channel that conducts current and permits control of the electron flow between the source and drain electrodes of the transistor. This layer is traditionally made of silicon dioxide but is now in transition to other, more complex, oxides that allow transistor size to be reduced even further. Therefore, silicon dioxide and other oxides with thicknesses of 20 angstroms or less are now of critical importance in the semiconductor industry.

The APXPS endstation at ALS Beamline 9.3.2 was the first and is still one of only a few places where real-time studies of these dielectric oxides can be carried out using photoelectron spectroscopy at up to a few torr of ambient pressure, which approaches the "normal" pressure levels of semiconductor processing. Researchers using this beamline were able to chart the rate of oxidation of silicon in the gate oxide layer for the first time. Understanding the rate of oxidation in this crucial layer will lead to better models for studying this layer and, ultimately, smaller transistors.

Silicon oxidation has been greatly studied via various characterization techniques. This research is generally divided into two categories: that which looks at oxide thicknesses of 100 angstroms or more, for which a D-G model can be used to predict the oxide growth kinetics, and more fundamental surface science studies at roughly the single monolayer level. However, a gap in our knowledge of kinetics exists in the thickness regime of current oxide gate layers, for which the D-G model does not apply. Real-time surface science studies of oxidation kinetics using photoelectron spectroscopy have also been limited to oxidation pressures of <10–5 torr, far below the multi-torr regime of real semiconductor processing.

Time evolution of the oxide thickness measured at an oxygen pressure of 1 torr and at oxidation temperatures of 450 ºC (green dots), 400 ºC (black dots), and 300 ºC (blue dots). The arrows indicate the approximate points of transition between the rapid and slow regimes. Red lines show least-square fits of the data in rapid and slow regimes.

The research team studied the oxidation rate of a Si(100) surface via chemical-state-resolved APXPS. They monitored Si 2p core levels in real time during oxidation, with spectra having high sensitivity to the surface and, via chemical shifts in binding energy, the individual oxidation states created during oxidation.

Left: A typical Si 2p core-level spectrum recorded at 350 eV photon energy from an ~22-Å-thick oxide on Si(100), together with a least-squares fit of various chemical components. Right: 3-D plots of a series of Si 2p core-level spectra taken at an oxygen pressure of 1 torr and at an oxidation temperature of 450 ºC.

The Si 2p core-level spectrum of oxidized Si exhibits a bulk-silicon component (Si0) and various oxide components that demonstrate a higher binding energy of up to about 4.4 eV. The peak at the lowest binding energy is accepted as that of elemental Si (Si0) and the peak at the ca. 4.4 eV higher binding energy as that of stoichiometric oxide (Si4+). From the measured ratio of Si4+ intensity to Si0, the thickness of stoichiometric SiO2 as a function of time can be calculated with 1–2 Å precision.

Si 2p spectra were obtained at 30-second intervals over a two-hour period at pressures of 0.01–1 torr and temperatures of 300–530 ºC. The time evolution of the oxide thickness reveals two kinetic regimes: a rapid regime at the beginning of oxidation and a subsequent slow regime in which the growth is much reduced. The oxide thickness of the break point dividing the two regimes is between about 6 and 22 angstroms and strongly depends on oxidation temperature and pressure. These drastic changes in rate curves cannot be explained by the D-G model. The oxidation rates in the rapid and slow regimes are 64–234 Å/h and 1.5–4.0 Å/h under test oxidation conditions, respectively, and are much larger than those deduced from the D-G model.

These rate curves should provide a unique test of alternative kinetic models in this important parameter regime. In addition to this study, oxidation of silicon with both oxygen and water present, as is often the case in industrial processing, has been examined, providing additional first-of-a-kind data for this important process. Future studies are planned at higher pressures and of more complex oxide combinations that are replacing simple SiO2.



Research conducted by Y. Enta (Hirosaki University, Japan); B.S. Mun (Hanyang University, Korea, and ALS); M. Rossi, and Z. Hussain (ALS); P.N. Ross Jr. (Berkeley Lab); C.S. Fadley (University of California at Davis and Berkeley Lab); and K.-S. Lee and S.-K Kim (Seoul National University, Korea).

Research funding: the U.S. Department of Energy; the Humboldt Foundation (Germany); the Helmholtz Association (Germany); the Jülich Research Center (Germany); the University of Hamburg (Germany); Overseas Advanced Education and Research Program from the Ministry of Education, Culture, Sports, Science, and Technology of Japan; Creative Research Initiatives of MOST/KOSEF (Korea).

Publications about this research: Y. Enta, B.S. Mun, M. Rossi, P.N. Ross, Jr., Z. Hussain, C.S. Fadley, K.-S. Lee, and S.-K. Kim, "Real-time observation of the dry oxidation of the Si(100) surface with ambient pressure x-ray photoelectron spectroscopy," Appl. Phys. Lett. 92, 012110 (2008); M. Rossi, B.S. Mun, Y. Enta, C.S. Fadley, K.S. Lee, S.-K. Kim, H.-J. Shin, Z. Hussain, and P.N. Ross, Jr., "In situ observation of wet oxidation kinetics on Si(100) via ambient pressure x-ray photoemission spectroscopy," J. Appl. Phys. 103, 044104 (2008).