Applying Rapid X-Ray Reflectometry to Advanced Interconnects

		At a Glance
	Rapid X-ray reflectometry (RXRR) is a new technique that provides rapid measurements of density, thickness and roughness of thin films and thin-film stacks required for advanced semiconductor processes.

As thin-film structures have become progressively thinner in advanced devices, control of thin-film thickness has become more critical.¹ At the same time, advanced processing techniques such as metal-organic chemical vapor deposition (MOCVD) and rapid thermal processing (RTP) have made other material properties such as density and surface roughness important factory metrics. Examples include thickness control of Cu seed/Ta (or TaN) barrier stack deposition in Cu metallization; process control for MOCVD barrier films; and control of porosity in low-k dielectric materials.

Traditional metrology techniques based on optical or opto-acoustic technologies encounter difficulties with these new processes, since they require accurate knowledge of material properties such as index of refraction or acoustic velocity to determine thickness. However, in many cases these material properties depend on the very process function to be monitored.

It has been known for years that X-ray reflectometry (XRR) is capable of determining thickness, density and roughness of thin films and thin-film stacks without prior knowledge of material properties.² In XRR, a monochromatic X-ray beam is reflected off a surface at near-grazing incidence (usually stepped between 0° and 2-4°). When the intensity of reflected X-rays is plotted as a function of incident angle, the reflectivity is unity at very low angles because all of the incident X-rays are reflected in the region of total external reflection. The reflectivity then falls off rapidly at a "critical angle" beyond which X-rays penetrate the material. If the material is a thin film, constructive/destructive interference between X-rays reflecting off the interface and surface lead to "fringes" on the reflectivity plot (Fig. 1).

1. With thin-film materials, constructive/destructive interference between X-rays reflecting off the interface and surface lead to "fringes" on the reflectivity plot.

It can be demonstrated from X-ray physics that the critical angle is dependent on wavelength, l, and the density of the film, r , while fringe spacing is dependent on l and film thickness, T; hence XRR can determine T and r independently. In addition, surface and interface roughness can also be determined from attenuation of the reflectivity curve and fringe amplitude. The technique is ideal for use with thin semiconductor films because it can measure films of <50 Å and up to ~2000 Å. Films in stacks of six or more films can be measured simultaneously, as long as there is a density difference of at least 10% between adjacent films.

2. The RXRR method illuminates samples and collects reflected x-rays from multiple angles simultaneously.

Unfortunately, because conventional XRR systems employ a "count-and-step" technique to obtain reflectivity vs. angle data, and because resolution on the order of 0.01° is required, several minutes are required to obtain XRR data. Therefore, the technique has not previously been compatible with a production environment.

A newly introduced XRR system — the Therma-Wave Meta-Probe X — solves this problem by illuminating the sample and collecting reflected X-rays from multiple angles simultaneously, utilizing a proprietary method employing a curved crystal monochromator and a position-sensitive detector (Fig. 2).³ This rapid XRR (RXRR) technique reduces the time per measurement to ~10 sec (rather than several minutes), enabling it to be used as a metrology tool for in-fab use.

Unlike optical or opto-acoustic metrology, the system has no moving parts, and the only calibration required is z-axis positioning of the wafer, which is easily accomplished via laser auto-focus.

Cu damascene processing

Because Cu is difficult to etch, a new method called damascene processing has been developed for Cu metallization. In the Cu damascene process, a thin barrier of Ta or TaN (250-400 Å) and a thin Cu "seed" layer (1000-1500 Å) are deposited on the patterned oxide surface in a PVD cluster tool. A thick (1 µm) layer of Cu is deposited on the seed-layer surface by electrochemical deposition (ECD). The excess Cu is then removed by chemical mechanical polishing (CMP), leaving the Cu wire structure on the polished surface.

3. The large fringes at high angle result from the underlying Ta, and the higher-frequency fringes at low angles result from Cu.

Control of the seed/barrier step is critical to the process, since the barrier must be thick enough to be conformal, but thin enough to minimize resistivity. The seed also must be conformal, but not so thick as to close off high aspect ratio features, leading to voids. Finally, because the Cu seed and Ta (TaN) barrier are usually deposited with advanced directional PVD techniques such as hollow-cathode PVD and ionized metal plasma (IMP) PVD, fairly frequent verification of film thickness and uniformity are necessary during production.

Figure 3 is an RXRR curve from a typical Cu/Ta wafer. The large fringes with the high angle result from the underlying Ta, while the higher-frequency fringes at the low angles result from Cu.

Because of the clear separation of the fringe sets, autofitting for thickness monitoring during production is straightforward. The entire data set is obtained in 10 sec, and the thickness and density of both films are determined simultaneously.⁴

4. Metrology techniques that probe surfaces at normal incidence require a micro-beam to measure field film thickness without interference from topology.

Although the area imaged on the wafer surface by X-rays is an ellipse ~2 × 7 mm, the RXRR technique can be used to measure thickness uniformity on product wafers. Metrology techniques that probe surfaces at normal incidence, such as optical or opto-acoustic tools, require a micro-beam to measure field film thickness without interference from topology (Fig. 4). Since XRR probes the surface at near-grazing incidence, the only X-rays that contribute to the specularly reflected beam are those that reflect from the top surface, while X-rays that are incident on trenches or vias are scattered out of the specularly reflected beam.

5. Radial line scans are shown from a patterned Cu seed (top)/Ta barrier (bottom) stack.

Figure 5 shows radial line scan data for a Cu/Ta coated patterned wafer, with alternating points from the center of dies and smooth "street" regions from the center to the edge of the wafer. This demonstrates that measurements from patterned areas can be used to determine deposition uniformity across a wafer. Similar data can be obtained as multipoint maps.

MOCVD barrier films

As metallization technology moves below 0.18 µm linewidths, thinner and more complex barrier structures will be required. Since MOCVD is essentially nondirectional, it is an attractive deposition alternative to PVD for high aspect ratio submicron features. Unlike PVD films, MOCVD films typically contain HC reaction products from the precursor gas, and therefore have lower density than PVD films. Tailoring purity is a major process challenge for MOCVD films, and measuring changes in density and roughness, as well as thickness, is a major metrology challenge.⁵

Figure 6 presents RXRR data for three MOCVD-deposited TaN barrier films, targeted at 50 Å thick and deposited at different reactor temperatures. The results showed a thickness range of 43-83 Å, and densities of 8.0-9.1 gm/cm³ (nominal density of PVD-deposited TaN is 14.8 gm/cm³). The lower densities are the result of HC residue from the MOCVD process.

6. Results showed a range of thickness of 43-83 Å and densities from 8.0-9.1 gm/cm3.

The density of the films can be tailored through post-deposition plasma processing. When an as-deposited film with thickness of 99 Å and density of 8.3 gm/cm³ is treated with the proprietary plasma process (Fig. 7), the resulting film is compressed to 78 Å and the density increases to 11.7 gm/cm³. Further processing reduces the thickness to 68 Å and increases the density to 12.2 gm/cm³, much closer to PVD TaN density.

Low-k dielectric films

Spun-on porous low-k dielectric materials have shown promise for reducing interlevel dielectric (ILD) constants to <2.5.⁶ The porosity of the film after processing determines the k value; therefore, measuring the porosity and cross-wafer porosity uniformity will be an important processing metric for these materials. Because the porosity is inversely related to the average density of the film, measurement of average density and knowledge of the bulk material density is sufficient to determine porosity.

The density of an as-deposited porous HSQ film was determined using the Meta-Probe X, and an average porosity of 58% was calculated. As is typical for these materials, a thin surface layer of higher density was observed on this sample.

Although higher porosity provides lower k value, structural integrity of the film is also an important parameter. Plasma processing of spun-on porous low-k materials has been found to reduce porosity and therefore increase mechanical properties, although it also increases the k value. When the as-deposited film was plasma-treated, the density was found to increase, and the calculated porosity decreased to 33%. RXRR data from a bare silicon wafer shows that both of the low-k materials had critical angles (hence density) much less than silicon.

Summary

7. Use of the proprietary plasma process reduced film thickness and increased film density.

RXRR is a new technique that provides rapid measurements of density, thickness and roughness of thin films and stacks required for advanced semiconductor processes. Examples have been presented of use in several advanced interconnect processing challenges, including Cu seed/Ta barrier control for Cu metallization; MOCVD TaN barrier film development; and porous low-k dielectric porosity control. By developing RXRR as a metrology tool, both process control and production metrology can be accomplished on blanket or product wafers.

William Johnson,general manager of Therma-Wave's X-Ray Products Group, has held several senior management positions during his 25 years in the industry. He received a bachelor's degree in physics and master's degree in electrical engineering from the University of Minnesota (Minneapolis), and a doctorate in materials science from Michigan Technological University (Houghton).

Lou Koppel,a senior physicist with Therma-Wave, has spent more than 30 years developing novel applications of X-ray technology, including the RXRR technology used in the Meta-Probe X. He received a bachelor's degree in engineering-physics from the University of California at Berkeley, and a master's degree in applied physics from the California Institute of Technology (Pasadena).

Tom Adamsis a writer and consultant based in Lawrenceville, N.J. He has written widely on semiconductor topics.

---

REFERENCES

1. A.C. Diebold, R.K Goodall, "Interconnect Metrology Roadmap: Status and Future," Proc. 1999 Intl. Interconnect Conf., 1999, p. 77.
   
2. D.K. Bowen, B.K. Tanner, "Characterization of Engineering Surfaces by Grazing-Incidence X-Ray Reflectivity," Nanotechnology, Vol. 4, 1993, p.175.
   
3. L.N. Koppel, L. Parobek, "Thin-Film Metrology by Rapid X-Ray Reflectometry," Proc. 1998 International Conf. on ULSI Technology, AIP Conf. Proc., Vol. 449, 1998.
   
4. W.C. Johnson, B. Relja, L. Koppel, S. Gopinath, "Semiconductor Material Applications of Rapid X-Ray Reflectometry (XRR)," Proc. 2000 Intl. Conf. on Characterization and Metrology for ULSI Technology, to be published.
   
5. A. Paranjpe, et.al, "CVD TaN Barrier for Copper Metallization and DRAM Bottom Electrode," Proc. 1999 International Interconnect Conf., 1999, p. 119.
   
6. E.T. Ryan, et. al, "Material Property Characterization and Integration Issues for Mesoporous Silica," Proc. 1999 Intl. Interconnect Conf., 1999, p. 187.
   

Talkback

We would love your feedback!

Post a comment

» VIEW ALL TALKBACK THREADS

Related Content

• Topics
• Author

Related Content

 

By This Author

There are no other articles written by this author.

SPONSORED LINKS