Gate process control using spectroscopic ellipsometry
04/01/2004
Metrology techniques such as atomic force microscopy and cross-sectional tunneling electron microscopy, while able to provide full 2D profile metrology with the requisite precision, have inadequate throughput to allow for proper statistical sampling necessary to control a volume manufacturing line. TEM has the additional drawback of being a destructive process. An alternative, optical CD metrology based on spectroscopic ellipsometry can be used instead. It is currently being utilized for inline process control and product disposition at the gate lithography and etch process steps on 130nm-generation logic devices manufactured in Texas Instruments' DMOS6 300mm wafer fab.
The ability to control the cross-sectional profile of polysilicon gate structures on semiconductor devices is paramount to maximize product yield and transistor performance. Tighter control of gate profile parameters leads to tighter distribution of transistor speeds, resulting in device performance that is optimized and consistent. Transistor performance is only fully confirmed during electrical testing after interconnect and metallization have been completed. The process steps that define the physical dimensions of the polysilicon gate, however, occur much earlier in the process flow.
Typically, there is a lag time of days or even weeks between the gate patterning steps and electrical test of the transistor. As a result, the ability to accurately measure physical profile characteristics of the gate inline at the patterning steps, and subsequently correlate those measurements to backend electrical test results, is critical for proper disposition of production material to maximize yield and minimize overall cost/good die. This importance is underscored by the significant increase in the number of chips on today's 300mm wafers.
For the past several years, the predominant metrology methods used for inline control of the gate process have been top-down measurement of the width (i.e., critical dimension, or CD) of the polysilicon gate using a low-voltage scanning electron microscope (CD-SEM); and electrical CD (ECD) measurement conducted via a parametric test system immediately after gate etch. Both methods have given precise measurement with relatively high throughput. However, both methods are limited in their ability to provide metrology or process characterization beyond the one-dimensional CD value.
Other metrology techniques such as atomic force microscopy (AFM) and cross-sectional tunneling electron microscopy (TEM) are able to provide full 2D profile metrology with the requisite precision. Throughput for these techniques, however, is inadequate to control a volume manufacturing line. Furthermore, TEM is a destructive process. Optical CD metrology based on spectroscopic ellipsometry — also referred to as SpectraCD — can provide 2D profile information on polysilicon gate structures.
Spectroscopic ellipsometry-based measurements
A typical setup for a spectroscopic ellipsometer is shown in Fig. 1. Light from a broadband source is reflected off the sample of interest. The polarizer and analyzer determine the state of polarization of incident and reflected beams, respectively. A prism separates wavelengths in the reflected light, and intensity is measured using an array detector.
 Figure 1. Schematic of spectroscopic ellipsometry-based CD measurement. |
To conduct a typical spectroscopic ellipsometry measurement, the intensity of the elliptically polarized light is measured for the wavelength range from 220–800nm, at different orientations of the polarizer or analyzer. Ellipsometry offers both amplitude and phase information, as opposed to reflectometry, which only provides amplitude information [1]. In this technique, the sample of interest is the grating targets added to the wafer, usually in the scribe line. The grating targets are comprised of line/space features of uniform period, with linewidth (CD) and period designed to represent the physical device feature being controlled.
 Figure 2. Measurement and library match using spectroscopic ellipsometry. |
A typical measurement process consists of an offline library generation process and an on-tool profile measurement using the library. The offline library generation uses the process information (dispersion properties and nominal thickness of all films in the grating region) and the grating information (pitch, nominal CD, HT) to create a theoretical model of the grating geometry. The library is a compilation of theoretical spectral signatures, which are obtained by varying grating parameters. This library is then linked to the recipe on the metrology tool. As the gratings are measured, the experimental spectra are compared against the theoretical spectra in the library (see Fig. 2). The best match between the measured and theoretical spectra determines the parameter values that best describe the grating [2].
Results
Results for 130nm-node gate lithography and etch processes are shown in Fig. 3. The same wafers and targets were measured a total of 30 times over a two-week period. Gratings with line-pitch ratios varying from 1:3–1:7 are used to monitor the lithography and etch processes. The aspect ratios on the lines in the gratings are ~2.3:1 for gate lithography, and ~1.6:1 for gate etch.
For both the gate lithography and etch processes, precision on CD and height parameters is well below 0.5nm and the precision on the sidewall angle is <0.1°. The resulting precision-to-tolerance ratios (P/T) are well below 0.1, providing sufficient capability to control a 130nm node process.
Correlation to drive current
One of the most important parameters for determining device performance is the transistor drive current (IDrive). IDrive cannot be measured until after metallization process steps have been completed. One of the primary contributors to determining IDrive is the physical gate length defined by the width of the polysilicon line (gate CD). Therefore, the ability to correlate a physical measurement of gate length measured inline at post-gate etch to IDrive, is critical to providing meaningful process control and product disposition that will maximize yield and minimize cost/good-die.
Historically, ECD measurements of gate CD have provided the best correlation to IDrive, yet some challenges have been encountered with maintaining this correlation at the 130nm process node. ECD measurement is completed immediately after the gate-etch process step using a parametric test system to measure a test structure in the scribe that simulates the actual device. While ECD provides relatively high throughput compared to other metrology techniques, it does face some limitations. There is motivation, therefore, to find a viable metrology alternative that provides better correlation to IDrive at higher throughput.
Figure 4 shows results of a study comparing the correlation of ECD and spectroscopic ellipsometry measurements to IDrive. The limitations of the ECD measurement are clearly seen. Spectroscopic ellipsometry shows a much higher level of correlation to IDrive (0.67–0.68) for both minimum contacted pitch (MCP) structures and isolated structures. The slope and correlation of the technique vs. IDrive demonstrate that IDrive can be controlled and modified through such inline measurements.
Process excursion detection
An additional benefit of spectroscopic ellipsometry is its capability to detect a variety of process excursions. The ability to generate full 2D profile information has provided additional sensitivity to process changes. The high throughput (<4 sec/site move-acquire-measure time) allows a significant increase in production sampling that leads to faster accumulation of statistically significant data, consequently reducing detection and response time for process excursions.
 Figure 5. Etch chamber post-PM excursion detection for a) SpectraCD gate-etch sidewall angle, and b) SpectraCD gate-etch CD. |
An example of detecting process excursions is shown in Fig. 5. After routine preventive maintenance was performed on an etch chamber, SPC charts for both sidewall angle and CD parameters reflected a 2% shift in the process means from the pre-PM baseline condition. Due to the high sampling rate, the excursion was reliably detected within a handful of product lots. TEM analysis confirmed the process shift. Follow-up maintenance was performed on the etch chamber to return the process to its baseline.
A unique capability of the measurement technique has been demonstrated: detecting gate-oxide punch-through, a yield-limiting defect that has proven difficult to detect in low levels using optical and/or CD-SEM inspection. In Fig. 6, very low levels of gate-oxide punch-through were detected by monitoring the goodness-of-fit (GOF) of the spectroscopic ellipsometry measurements. When the measured spectrum from a grating target is compared against the theoretical library, the technique generates a GOF value that mathematically represents how closely the measured spectrum matches the solution from the library. The small holes in the gate oxide underneath the grating target scatter the incident spectroscopic ellipsometry beam, introducing random noise into the measured spectra, which leads to a decrease in GOF.
 Figure 6. GOF plot with corresponding SEM images confirming detection of gate-oxide punch-through. |
The plot in Fig. 6 shows a clear decrease of 0.02–0.06 in GOF from a mean value that is typically stable to within ±0.01. Subsequent CD-SEM inspection of the grating targets highlighted the small holes representing the gate-oxide punch-through. As a result of the reliable detection method, an inline CD-SEM inspection has been eliminated, saving cost without sacrificing detection of punch-through defects and proper disposition of production material.
Conclusion
The long-term results for the gate lithography and etch structures show that spectroscopic ellipsometry is capable of meeting the precision requirements for the 130nm node as outlined in the International Technology Roadmap for Semiconductors. Measurements of the grating target in the scribe line show good correlation with the transistor drive current, and therefore provide an inline measurement technique at the gate etch step that can be used to perform effective process control and product disposition. Additional value has been recognized with the technique's ability to quickly detect process excursions. Because of its performance, the technology has been implemented at Texas Instruments' DMOS6 production facility for inline process control and product disposition at the gate lithography and etch process steps for 130nm-node volume production.
Acknowledgments
The authors wish to acknowledge Dale Burrows and Ray Chiao from Texas Instruments, and Srini Rangarajan, Kamal Bhatia, and Suresh Lakkapragada from KLA-Tencor for their contributions to this article. SpectraCD and SCD are registered trademarks of KLA-Tencor.
References
1. H. Tompkins, W. McGahan, Spectroscopic Ellipsometry and Reflectometry, John Wiley & Sons, 1999.
2. J. Allgair, D. Benoit, M. Drew, R. Hershey, L. Litt, et al., "Implementation of Spectroscopic Critical Dimension (SCD) for Gate CD Control and Stepper Characterization," Proc. SPIE, Vol. 4344–57, March 2001.
For more information, contact J. Scott Hodges, Plasma Processing, at Texas Instruments Inc., 13011 T.I. Blvd., Dallas, TX 75243; ph 972/995-8912, e-mail [email protected].
Yu-Lun (Chris) Lin is a plasma processing engineer at Texas Instruments Inc.
Robert Peters is applications project manager at KLA-Tencor Corp., 1717 Firman Dr., Suite 100, Richardson, TX 75081; ph 972/664-8336, fax 972/690-0345, e-mail [email protected].