RequestLink
MICRO
Advertiser and
Product
Information

Buyer's Guide
Buyers Guide

tom
Chip Shots blog

Greatest Hits of 2005
Greatest Hits of 2005

Featured Series
Featured Series


Web Sightings

Media Kit

Comments? Suggestions? Send us your feedback.

 

MicroMagazine.com

Defect/Yield Analysis and Metrology

Using an XPS-based metrology system to determine film thickness and composition

C Thomas Larson, Emir Gurer, and J. Kelly Truman, ReVera

The IC industry is experiencing a major transition. Mobility and integration have become the drivers of semiconductor products, replacing computational speed as the primary growth factor. New features determining design are the number of integrated functions, portability, and battery-life hours. Devices such as cell phones integrated with cameras, wireless cable setup boxes, and plasma TVs are propelling chip scaling as much as new network servers and workstations.

It is well understood that new materials, processes, and tools must be developed to satisfy the growing demand for powerful, portable chips. In the process-tool area, the metrology segment is growing faster than any other group. It is becoming increasingly true for devices at the 90-, 65-, and 45-nm nodes and below that process latitude is shrinking as new materials are being integrated. Established processes are being pushed until the last bit of margin can be squeezed out of them. At the nanometer level, process tools are being forced to operate at the atomic scale.

Metrology must be increasingly deployed in fabs to enable faster process development, help put inoperable process tools back on-line quickly, provide a mechanism to ensure that duplicate process tools are well matched, and—most importantly—detect process excursions. In the past, metrology has been focused solely on the dimensions of a target—its width, depth, or thickness. That is changing, however. Metrology must adapt as final device performance in a growing number of process areas is being determined not by the size of a feature but by how many atoms of a specific type are in it—in other words, by its composition.

The emergence of atomic-scale metrology is predictable. In the 1970s, labs used reflectometers to measure film thicknesses, while fabs used visual color charts based on process margins. In the 1990s, labs used x-ray technologies to measure very thin films and determine film composition, while fabs, still relying on wide production margins, only controlled thicknesses. In the nanometer era, however, tiny atomic-scale process margins require that x-ray techniques, which used to be confined to the laboratory, must be automated and integrated into the fab.

Atomic-Scale Metrology

Atomic-scale metrology began to appear with the advent of 90-nm logic technology. Starting with transistor processes, leading-edge (65- and 45-nm) logic processes require atomic-composition process control steps in addition to traditional thickness measurements. Likewise, DRAM and nonvolatile memory applications rely heavily on complex films to achieve ever-increasing memory densities. DRAM products face the toughest challenge because of changing transistor and capacitor materials. In copper processes, the likely switch to atomic-layer barrier deposition at 45 nm will require chipmakers to control metal nitrides 6–9 nm thick and films composed of various materials.

Advanced metrology is most crucial in the transistor gate dielectric process. While silicon dioxide has been used traditionally in this process since the advent of CMOS, the intrinsic inability of SiO2 to maintain necessary electrical properties as the technology is scaled has led to the introduction of a new atom into the SiO2 matrix—nitrogen. Between 1 and 2 nm thick, a state-of-the-art SiON film contains between 5 and 25% nitrogen, depending on the device design rules. As the transistor gate dielectric process has evolved from a thermal diffusion process to include various plasma-enhanced deposition processes, metrology needs have shifted as well.

A 2-nm-thick film is about eight atoms high. At that scale, thickness is difficult to define, let alone to measure. With nitrogen comprising one to three of those atoms, it too must be controlled. While this film has traditionally been measured using optical film-thickness metrology, the introduction of nitrogen has changed the metrology requirements.

Figure 1: Correlation between the change in nitrogen concentration of the SiON film and the dielectric constant and EOT.

Figures 1–3 demonstrate the difficulty of controlling compositionally complex processes. Figure 1 shows the dependence of the SiON dielectric constant on nitrogen concentration. In this case, the dielectric constant is represented as the ratio of the physical film thickness to the electrically equivalent oxide thickness (EOT). Ultimately, EOT is the variable that is changed by adding nitrogen. The gate dielectric process, using nitrogen, increases the dielectric capacity of the gate oxide while maintaining a physically thicker film. However, the film behaves electrically as though it were thinner. Figure 1 illustrates that any change in the nitrogen concentration of the film causes a proportional change in the dielectric constant and EOT. Therefore, if the nitridation process drifts, it causes a drift in the electrical performance of the device.

Figure 2: Schematic diagram showing the formation of SiON gate dielectric with the addition of nitrogen.

To form a SiON gate dielectric, as shown in Figure 2, a starting film of SiO2 is created on a silicon substrate. After the SiO2 is formed, nitrogen is added. Traditionally, in order to keep the gate dielectric process in control, SiO2 film thickness is monitored. However, because the critical electrical properties of the film strongly depend on nitrogen concentration, it too must be measured.

Finally, Figure 3 illustrates the impact of controlling only film thickness. In this case, a <2-nm SiON film was created. If film thickness alone had been measured, the film would have presented no obvious yield problems, as indicated in Figure 3a, in which thickness uniformity was found to be <1%. But at the first electrical test, a significant within-wafer electrical nonuniformity was revealed, as shown in the nitrogen map of the film in Figure 3b, in which nitrogen concentration uniformity across the wafer was >8%.

Figure 3: Wafer maps illustrating (a) that final film thickness with a uniformity of <1% at 15 Å is in control, and (b) that the final nitrogen distribution with a uniformity >8% at 1.0 X 1015 atoms/cm2 is out of control.

X-Ray Photoelectron Spectroscopy

The measurements presented in Figures 1–3 were made using an RVX 1000 in-line compositional metrology system from ReVera (Sunnyvale, CA). The tool is based on x-ray photoelectron spectroscopy (XPS), which has been used in the laboratory for more than 50 years and is emerging as a reliable method for measuring the thickness and composition of semiconductor ultrathin films. Two main features distinguish the semiconductor metrology tool from its laboratory cousin: its production-worthiness and precision. Precision at a useful throughput is needed to ensure adequate process control, and round-the-clock operation is required in today's fabs.

Figure 4a presents a schematic diagram of the XPS concept. Low-energy x-rays illuminate a surface and penetrate to a few micrometers in depth. When an x-ray is absorbed by the surface atoms, the atoms must conserve energy. In many cases, the atom ejects an electron. The electron is ejected with energy equal to the initial kinetic energy of the illuminating x-ray plus the energy that bound the electron to the atom. Ejected electrons are called photoelectrons.

Figure 4: Schematic diagram showing (a) emission of photoelectrons from a film surface, and (b) typical compositional spectrum from an SiON surface with several photoelectrons' binding energies.

The photoelectrons are collected and analyzed according to their energy. Because of the low energy of the illuminating x-rays, only photoelectrons near the surface are collected. Photoelectrons from deep within the material are reabsorbed. The XPS method can generally detect photoelectrons approximately 10 nm from the top surface of the film.

Figure 4b shows a typical compositional spectrum from an SiON surface with several important binding energies. The first is the binding energy for silicon at approximately 100 eV. The inset indicates that the silicon photoelectron energy actually has different components. The spectroscopic information is composed of the outer-shell electrons of the measured material, and XPS is sensitive enough to differentiate among the material's bonding states. In this case, the lower- binding-energy peak corresponds to photoelectrons emitted from the the silicon substrate (Si-Si) bond, while the higher- binding-energy portion of the spectrum corresponds to photoelectrons emitted from silicon atoms inside the SiON film itself, whose binding energy increased because of its chemical structure (Si-O and Si-N bonds). Most of the silicon atoms are bound to oxygen, but some are bound to nitrogen.

The figure also shows the relative amounts of important elements in the film. Nitrogen, oxygen, and—interestingly—carbon are of interest. The carbon may come from the process or an adsorbed layer on the film surface. Such adsorbed organic layers are typically referred to as molecular atmospheric contamination (MAC) and are caused by amines and other airborne contaminants. In evaluating the composition and thickness of the film, the MAC layer can be disregarded, since its effect is in the spectral region of carbon and not in the energy regions of interest.

Figure 5: Schematic diagram showing (a) that when x-rays pass through a surface, Si-O and Si-Si photoelectrons are emitted, and (b) Si-O and Si-Si peaks. Thickness is determined by measuring the attenuation of one type of photoelectron as it passes through a film above it.

Figure 5 illustrates another feature of the x-ray technique. XPS thickness measurements are made by comparing peak ratios. The general theory governing this method is called attenuation theory. Thickness is determined unambiguously by measuring the attenuation of one type of photoelectron as it passes through a film above it. For example, in the case of SiON on silicon, photoelectrons from the silicon substrate are measured by evaluating the peak of the Si-Si bond. When an SiON film is on top of the silicon, Si-Si photoelectrons must pass through that film to be detected. As SiON film thickness increases, fewer Si-Si photoelectrons pass through the film. Hence, by calculating the ratio of the Si-O peak to the Si-Si peak, the thickness of the film can be determined. The attenuation theory enables investigators to determine the thickness of multiple film stacks, provided their total thickness is in the range of 10 nm. The literature provides a good overview of this theory and its relationship to thickness measurement via XPS.1

XPS is conceptually simple, nondestructive, and analytically powerful. All composition and thickness information is contained in the spectral output of the photoelectron energy analyzer system. The number of photoelectrons detected at a given energy is proportional to the number of atoms of that type in the surface. In the lab, a surface scientist is typically required to operate the tool and interpret the results. However, in order for the XPS technique to qualify as production worthy, fab personnel not intimately familiar with the underlying technology must be able to operate it. Just like its laboratory predecessors, the reflectometer and the ellipsometer, XPS has emerged from the lab enhanced and optimized for use as an in-line metrology system to control process variations associated with atomic-scale thickness and composition.

Other analytical techniques that provide compositional information can generally probe only the inner, core electrons of atoms. Chemical bonding information cannot be obtained from most electrons because of the electrons' very high binding energies.

Results Using In-Line XPS

An example of how in-line ultrathin film compositional metrology can be applied is shown in Figure 6 and Table I. The figure presents typical wafer maps from an SiON process that was altered to improve process tools' particle generation performance and to accommodate a new device design. While the process tool modifications resulted in reduced particle levels, they unexpectedly caused compositional variations in the nitrogen concentration.

Figure 6: Typical wafer maps from an SiON process showing (a) and (b) film thickness and nitrogen-dose maps before a process tool modification, and (c) and (d) film thickness and nitrogen-dose maps after the modification. While the process tool modifications resulted in reduced particle levels, they caused compositional variations in the nitrogen concentration.

Historically, engineers would have made the tool modifications and verified the results using thickness metrology only. In this case, however, they were saved from having to perform weeks of troubleshooting work on a degraded process. Using the in-line XPS tool, further adjustments and modifications enabled the device maker to satisfy both film uniformity and particle requirements.

Process Condition
Film Thickness (%)
Nitrogen Dose (%)
Before tool modification
0.16
1.35
After tool modification
0.19
2.79
Table I: Thickness and nitrogen dose uniformity before and after process tool modification.

Gate processes are not the only compositionally critical ones in semiconductor manufacturing. The new DRAM and flash devices use dielectrics that are composed of hafnium oxide alloys. Much work is being conducted on the integration of some form of aluminum and hafnium oxide. Evidently, aluminum improves leakage performance and stabilizes HfO2 thermally. However, the use of aluminum in terms of device performance reduces the overall capacitance and, therefore, potential memory density. Hence, a careful balance must be struck to manufacture such devices optimally.

In developing typical ALD films composed of hafnium and aluminum, engineers must be able to measure process uniformity across the wafer and achieve compositional process repeatability, since stoichiometry is critical to determining capacitor performance. Figure 7 shows thickness uniformity for three different high-k capacitor films (HfO2, Al2O3, and SiO2) and compositional uniformity (in %) for hafnium, aluminum, and oxygen. In a production setting, the ability to monitor the uniformity of the relative ratios of hafnium to aluminum or oxygen is crucial for achieving optimal yields.

Figure 8: Compositional metrology results showing the correlation between residual chlorine and surface resistivity.

In addition to its application in the capacitor process, titanium nitride is beginning to appear in the electrodes used in some DRAM architectures. Sometimes this film is created from a CVD process based on chlorine precursor chemistry. While generally a well-understood process, the use of this chemistry leaves residual chlorine on the surface, which can cause the electrode to exhibit poor resistivity. As illustrated in Figure 8, compositional metrology was used to demonstrate the dependence of surface resistivity on residual chlorine. These data show that even a 1% change in chlorine on the surface doubles resistivity. Hence, proper control of chlorine on the wafer surface is critical.

The high-k films that have been developed for memory and other hafnium systems, which are typically ALD films composed of nitrogen-doped HfSiO or HfO2, are also being contemplated for future generations of the gate-oxide process. The XPS metrology tool has been qualified and is being used to develop and control these next-generation films.

The ALD process generally requires a specific surface condition to enable the deposition reaction cycle to saturate. The first few ALD cycles are key. ALD works by exposing a surface to a sequential and alternating set of precursors; at each cycle, it lets the surface reaction saturate, after which no further film deposition occurs. If complex films are required, alternating exposures to several different precursors are performed.

In the HfO2 gate dielectric case, the initial surface state of the silicon is the key to success. Because of the reactions used, there is always an interfacial layer of SiOx (suboxide). As a result, a two-stack film with a high-k film (HfO2) in series with a low-k film (SiO2) is formed. If the optimal suboxide is not formed, the resulting electrical characteristics are degraded. Figure 7 shows how compositional metrology can be used to determine the thickness and composition of the HfO2 and suboxide layer simultaneously.

Figure 9: Compositional maps illustrating the concentration ratio of silicon to germanium in the SiGe film.

Several other critical front-end of-line processes require compositional control. For example, silicon germanium depositions are used increasingly to form the strained-silicon channels for high-performance logic devices. While these films are typically at least 100 nm thick, the top surface is where the action is. For example, Figure 9 indicates that by measuring the silicon-to-germanium concentration ratio at the surface, users of the XPS system can control the processes typically used to deposit the film. Temperature and gas-flow stability are crucial in these processes, and any process drift manifests itself as a variable ratio of silicon to germanium. That variation affects the strain capacity of the film.

Another very important process under investigation is shown in Figure 10. The addition of a specific concentration of fluorine to the transistor process may help offset poor neutral bias thermal instability (NBTI) performance. NBTI manifests itself as a long-term drift in the threshold voltage. Occurring more frequently with heavily nitrided gate oxides and high-dose ultrashallow junctions, it is not a conveniently reversible process. The addition of fluorine tends to stabilize the phenomenon.

Figure 10: Compositional map of fluorine distribution.

Conclusion

The future will see many new composition-critical films. According to the International Technology Roadmap for Semiconductors, future ultrashallow junctions will be formed at depths of 10 nm or less, with dopant concentrations surpassing the solid solubility limit of silicon. This establishes a clear need for new metrology methods to control very critical implant process steps. Traditional in-line control metrologies will fail to detect implanter dose excursions, because they cannot measure the presence of dopants that are not activated. True physical compositional metrology is required. At high doses and low implant depths, XPS-based technology provides an excellent mechanism for monitoring implant dose.

At the back end of the line, where the copper dual-damascene process has been used since the 0.18-Ám technology node, most yield-limiting process problems have been physical failures resulting from integration defects such as voids and delamination. In the coming device generations, surface conditioning will enable more-robust integration, so that the chemical state of a surface will determine its yield. For example, a low-k surface will be prepared for subsequent barrier deposition and an ultrathin TaN barrier will be prepared for optimal electroless copper deposition. These process steps will require new methods of process control. Because of their strong surface sensitivity, XPS-based technologies will be a strong choice.

At the same time, process control of film deposition will continue to use traditional metrologies at an increasing rate. Last year, VLSI Research projected that the overall film metrology market will reach approximately $900 million in 2009. Although not explicitly tracked by the major market agencies, the segment devoted to compositional metrology is projected to occupy a significant part of that market. Many XPS-based compositional metrology tools are used in advanced fabs around the world. With an increasing installed base, the technology has arrived as a standard tool for process control applications.

Acknowledgments

The authors wish to thank Kathy Barla, Jerome Bienacel, and Nicolas Emonet from the Crolles2 Alliance in Crolles, France, for the data appearing in Figures 1 and 8. They would also like to thank Zoe Osborne, Don Wayne, and Mike Kwan from ReVera's applications engineering team.

Reference

1. D Briggs and MP Seah, Practical Surface Analysis, vol. 1, 2nd ed. (New York: Wiley, 1990).


C. Thomas Larson is vice president of product marketing at ReVera (Sunnyvale, CA), which emerged out of the metrology division of Physical Electronics in 2003. He came to Physical Electronics from Applied Materials, where he was director of marketing for the advanced process control and integrated metrology programs. He was also involved in strategic and field marketing at Schlumberger, where he developed and supported worldwide field operations for E-beam inspection, CD SEM, and overlay registration tools. Larson began his career in capital equipment at KLA-Tencor, where he held applications and product-management positions in the Surfscan and other strategic units. He received a BS in physics from Arizona State University in Tempe and performed postgraduate work in physics at San Francisco State University. (Larson can be reached at 408/205-9235 or tlarson@revera.com.)

Emir Gurer, PhD, is director of applications engineering at ReVera. He came to the metrology division of Physical Electronics from ASML's track division, where he was director of process technology. His responsibilities at ASML included advanced process technology and applications development, field applications, and new product development. Before joining ASML, Gurer was a postdoctoral research associate at the Zettlemoyer Center for Surface Studies at Lehigh University in Bethlehem, PA. He holds eight patents and has presented and published more than 25 papers in national and international conferences. He received a BS and MS in physics from METU in Ankara, Turkey, and an MS and PhD in physics from Lehigh University. (Gurer can be reached at 408/530-3828 or egurer@revera.com.)

J. Kelly Truman, PhD, is vp of strategic marketing and technology at ReVera. Before joining the company, he served as general manager of the wet-clean division at Applied Materials, where he had previously been responsible for product and strategic marketing for the start-up of the rapid thermal processing group. Also before joining ReVera, he spent a year on the materials science faculty at the University of Florida in Gainesville, where he investigated nano/biosensor materials while consulting for major semiconductor capital equipment companies. He has also held technical positions at Conductus, CVC, Tektronix, and IBM. He received a BS in metallurgical engineering from the University of Notre Dame in South Bend, IN, and MS and PhD degrees in materials science and engineering from the University of Florida. (Truman can be reached at 408/530-3858 or ktruman@revera.com.)


MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit

Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com.

© 2007 Tom Cheyney
All rights reserved.