Laser ellipsometry (LE) and DUV reflectometry (DUVR) are now being used to monitor etch processes and reduce nonuniformities that had affected yield. Maps showing 49-point thickness and reflectance of an ARC/polysilicon/oxide/silicon stack before process optimization are presented to illustrate the way such data are used for in-line monitoring of etch performance and to control and determine etch characteristics .
Extensions to optical lithography have exceeded all expectations. Manufacturers now routinely pattern features that are approximately half the size of the lithography wavelength. A variety of different materials and techniques are used to create 130nm features with 248nm wavelengths and will soon be widely implemented to create 90nm features with 193nm wavelengths. To maintain high-yield at these challenging dimensions, every step of the lithography process requires strict control.
Antireflective coatings (ARC) are one of the key enablers for advances in lithography, improving CD control by reducing or eliminating reflections from the surface or underlying layers that can result in uneven photoresist exposure. The challenges facing ARC designers increase significantly with the shorter wavelengths required for patterning narrow lines. With deep-ultraviolet (DUV) wavelengths, very small changes in photoresist thickness can result in large resist sensitivity changes due to thin film interference effects. Also, the reflectivity of the substrates often increases dramatically .
To control these effects, manufacturers are likely to use a variety of ARCs during production that can be deposited on the surface of the photoresist or on the substrate. ARCs may be spin-on organic materials, which are stripped with the photoresist, or inorganic SiOxNy materials that may be integrated into the final devices. SiOxNy ARCs have some advantages over spin-on materials:
- The optical values (n and k) can be tuned (i.e., tunable ARCs) for the application and for the photoresist being used by controlling deposition parameters .
- The CVD process is fast and provides a very conformal coating with a thickness that can be precisely controlled.
- The SiOxNy layer may act as a passivation layer, hard mask, or etch/polish stop in other processing steps.
Tunable ARCs are required to help control critical applications, including gate definition and contact opening, and are necessary for high-performance, high-yield production.
Process control requirements
When developing a new process, manufacturers will precisely tune SiOxNy ARC thickness and stoichiometry to meet specific needs. These values must be carefully maintained as the process is ramped to high volume. Measuring optical constants can determine if the deposited film has the correct composition for desired antireflective, mechanical, and electronic properties. The film must be maintained at the target thickness to prevent both over-/under-etch and over-/under-polish, and to act as an effective ARC. Small variation in film stoichiometry and thickness across the wafer, or from wafer to wafer, can lower device performance and reduce yield.
From CVD chamber to chamber, thickness uniformity of better than ±3% of the target is required, with even tighter single-chamber requirements. For a film thickness of 500Å, this translates into a process tolerance of 30Å. For metrology, the maximum precision/tolerance (P/T) is 0.3. As can be seen in the equation, this translates into a required metrology reproducibility and tool matching of better than 1.5Å for process control.
Using a similar process, the long-term metrology reproducibility and matching for the index of refraction, n, must be >0.0015.
It is difficult for spectroscopic ellipsometers to achieve this level of performance because of the instability of the white light source and complications arising from wavelength calibration and matching; however, manufacturers have a critical need for a metrology that can adequately control their CVD equipment across a fab or from fab to fab.
A combination of LE and DUVR can meet or exceed the metrology reproducibility and tool-to-tool matching required for ARC process control previously described. Laser ellipsometry is ideal for measuring the thickness of these very thin, sub-500Å films. The optical properties must be measured at the lithography wavelengths of 193nm or 248nm, where most systems are light-starved. The Rudolph DUV reflectometer uses a photomultiplier detector to provide sensitive detection. This technique is suitable to high-volume manufacturing needs, as thickness, n, and k can be measured simultaneously in seconds.
Figure 1. Measured and best-fit modeling results for a silicon oxynitride ARC with a) 633nm ellipsometry and b) DUV reflectometry.
Figure 1 shows results from a typical SiON ARC. Figure 1a shows the ellipsometric measurement of delta and psi at 633nm. Figure 1b is the reflectance of the same film measured from 190–360nm. The excellent fit between the measured and best-fit modeling results enables the accurate modeling of thickness and optical constants.
Stable laser light sources and high-intensity DUV reflectometry provide the kind of reproducibility and tool-to-tool matching shown in Table 1. These results were obtained by taking three measurements/day over a five-day period on two different metrology tools. The average thickness was measured to be 483.273Å with a standard deviation of 0.168Å on tool A, and 482.531Å with a standard deviation of 0.116Å on tool B. The two tools matched to 0.742Å. The optical constants n and k at both 248nm and 193nm are also shown in Table 1. All of the measurements exceed the reproducibility required for process control of this ARC. This level of tool-to-tool matching has been achieved on more than ten tools across multiple fabs, enabling manufacturers to match process equipment across a fab and around the world.
SIS application and results
The accuracy of the LE/DUVR thickness measurements was recently confirmed in an ARC application at Silicon Integrated Systems (SIS). In the SIS process, an ARC layer is deposited on polysilicon; in subsequent steps, cobalt is deposited on the polysilicon to form cobalt silicide. It was found that nonuniformities in the ARC layer were causing a rough cobalt silicide/silicon interface with a higher-than-expected resistance, thus requiring the ARC etch process to be monitored.
Accuracy was critical in this application, so TEM confirmation of the results was required. Wafers were prepared for three different stacks: 1) ARC/polysilicon/oxide/silicon, 2) ARC/oxide/silicon, and 3) ARC/silicon.
The ARC (SiOxNy) layers were formed by CVD and then etched for three different time periods: 0 sec, 180 sec, and 360 sec. The approximate thickness of the ARC layer at each of these times was ~250Å, 150Å, and 75Å, respectively. A top layer of polysilicon was added to each stack for the TEM cross-sectioning.
Figure 2. Simultaneous measurement of the thickness of the overlying polysilicon (4T), the ARC layer (3T), and the volume fraction (2Vf) and thickness (2T) of the underlying polysilicon.
LE/DUVR models were developed to measure the thickness of the ARC material through the polysilicon on all three stacks. For the first stack, the thickness of the top polysilicon and the thickness and amorphous silicon volume fraction (Vf) of the underlying polysilicon were measured simultaneously with the ARC thickness. A typical measurement is shown in Fig. 2, along with the TEM of the ARC thickness. Before capping, the ARC had been etched for 180 sec. The system measured the overlying polysilicon to be 603.30Å (4T), the ARC layer to be 150.18Å (3T), the Vf of the underlying polysilicon to be 14.997Å (2Vf), and the thickness of the underlying polysilicon to be 1963.8Å (2T). This matches very well with the TEM measurements for the thickness of the ARC, which is shown in three sites to measure 150Å, 154Å, and 157Å.
Table 2 compares TEM measurements to LE/DUVR metrology measurements at all three etch times for the first stack. At 0 sec of etch, TEM results measured an average thickness of 249Å, while the metrology system measured 246.61Å. For 180 sec and 360 sec, the TEM measured 153Å and 77Å, respectively, while the system measured 150.18Å and 76.29Å. Similar results were achieved on polysilicon/ARC /oxide/silicon and polysilicon/ARC/silicon stacks, confirming that LE/DUVR thickness measurements are very accurate for all three stacks.
Figure 3. Map with 49-point comparison of thickness and reflectance at etch times of a) 360 sec and b) 480 sec before process optimization.
With demonstrated accuracy and reproducibility, SIS now uses the LE/DUVR to monitor its etch process and to reduce nonuniformities. Maps showing 49-point thickness and reflectance of the ARC/ polysilicon /oxide/silicon stack before process optimization are shown in Fig. 3. At 360 sec, the ARC layer was generally center-thick and edge-thin with the thickness ranging from a minimum of 66Å to a maximum of 120Å. Reflectance at 248nm ranged from 33% min. to 52% max. At 480 sec, the ARC layer had almost been cleared. The thickness ranged from 19Å min. to 25Å max., and the reflectance has increased significantly, ranging from 62% to 63%. Detailed information such as this is used for in-line monitoring of etch performance and to precisely control and determine etch characteristics.
SiON ARCs are a critical element for high-yield lithography. But maintaining process control of CVD chambers requires very reproducible metrology measurements with excellent tool-to-tool matching. The LE/DUVR metrology system meets these requirements and is designed to monitor high-volume production. At SIS, the capability of the LE/DUVR system to accurately measure the thickness of very thin SiON ARC films was tested. When compared to TEM results of films ranging in thickness from 250Å to 75Å, the LE/DUVR showed excellent correlation, and therefore is now being used to monitor yield-critical ARC etch applications to improve uniformity.
- G. Jiang, et al., "Characterizing SiOxNy ARC Materials with Laser Ellipsometry and DUV Reflectometry," Proc. SPIE, Metrology, Inspection, and Process Control for Microlithography XVIII, Vol. 5375, 2004.
- R.R. Damel, "Anti-Reflective Coatings: Theory and Practice," SPIE Education Services — Microlithography 2003, Feb. 2003.
- K. Suhm, et al., "Making 50nm Contact Holes with DUV," Microlithography World, Vol. 12, No. 3, pp. 4–6, 19.
Gary Jiang is an advanced application engineer at Rudolph Technologies, One Rudolph Rd., Flanders, NJ, 07836; ph 973/448-4494, fax 973/691-4863, e-mail email@example.com.
Jui-Ping Li is manager of the Die Production/Diffusion Engineering department at Silicon Integrated Systems, Taiwan, ROC; ph 886/3-5790168 ext. 3240, fax 886/3-5630449, e-mail firstname.lastname@example.org.