- Biomedical Optics & Medical Imaging
- Defense & Security
- Electronic Imaging & Signal Processing
- Illumination & Displays
- Lasers & Sources
- Micro/Nano Lithography
- Optical Design & Engineering
- Optoelectronics & Communications
- Remote Sensing
- Sensing & Measurement
- Solar & Alternative Energy
- Sign up for Newsroom E-Alerts
- Information for:
Sensing & Measurement
Hybrid metrology for advanced semiconductor fabrication
A next-generation measurement solution for complex semiconductor devices may find a way around the limitations of current techniques.
17 August 2011, SPIE Newsroom. DOI: 10.1117/2.1201108.003827
“If you cannot measure it, you cannot improve it,” in the words of Lord Kelvin (1824–1907). Consequently, as semiconductor feature sizes and process tolerances shrink, there are greater requirements for speed in metrology (the science of measurement), as well as precision and measurement accuracy. For example, critical dimension (CD) metrology for advanced semiconductor chips must be correct well below the size of a single silicon atom (0.25nm). Automated in-line measurements are necessary to provide control in semiconductor processing and ensure that chips are manufactured within requirements for yield, variability and performance.
Industry typically tries to solve the challenges posed by in-line measurement of semiconductor devices by focusing on an individual toolset such as scatterometry (optical CD), CD scanning electron microscopy (CDSEM), and CD atomic force microscopy (CDAFM). These use different probes and measurement methods for the same measurand and each has inherent limitations, merits and assumptions. For instance CDSEM is limited to top-down measurements. OCD offers a solution for measurement of profile details but suffers from difficulties with inter-parameter correlation. In AFM, the probe size and shape limit the ability to physically scan small trenches or measure accurately near the bottom region of structures. ‘Hybrid metrology’ brings together the strengths of two or more metrology toolsets to provide more comprehensive measurement of the same measurand than these techniques can individually provide.1–4 Data obtained from one tool is shared with another tool and used in a complementary or synergistic way to enhance the resolving power of either or both tools.
In recent work,1we have demonstrated how hybrid metrology can help alleviate some of the key concerns faced by semiconductor metrology. These include accurate measurement of 3D structures such as fin-shaped field effect transistors (FinFETs), metrology robustness and stability, and the length of time taken to set up the metrology process.
Simple hybrid metrology schematic.1
CD: Critical dimension. Thk: Thickness. N&K: Optical constants.
There is more than one way of implementing hybrid metrology in semiconductor fabrication (see Figure 1for one approach). When combining data from multiple metrology tools, the secondary or source toolset is defined as the tool that generally measures first and provides the information to the primary or receiver toolset. Data from the secondary tool serves two purposes: it improves the measurement performance of the primary tool by providing additional data; and it provides reference system feedback to ensure stable metrology performance.
Hybrid metrology of a complex 3D structure. (a) Fin-shaped field effect transistor (FinFET) gate final inspect schematic. (b) Hybrid scatterometry (optical critical dimension, OCD) measurements (right) allow greater accuracy than using individual OCD measurements (left).1
Scale: 2nm increments on both x and y axes. Ht: Thickness. TEM: Transmission electron microscopy. GOX +HK: Thickness of gate oxide and high-k (high dielectric constant) films. OCD: Optical CD metrology (scatterometry). Ox: Oxide. Poly: Polysilicon film. R2
: A statistical measure of how well a regression line approximates real data points. SiN: Silicon nitride. SWA: Sidewall angle. TMU: Total measurement uncertainty.
We have demonstrated a notable improvement in measurement performance by using hybrid metrology on complex 3D structures, including at the FinFET gate etch process step: see Figure 2(a) for a schematic of the process. Due to the complexity of this structure, there are many possible ‘floating’ parameters in the OCD model. This typically leads to the issue of correlation between multiple parameters and a lack of sensitivity in measurements of the desired parameter. As a consequence there is high measurement noise on critical parameters, which in turn could result in high variability among these parameters on product wafers. To evaluate our hybrid metrology, we made in-line measurements of the geometrical parameters metal CD and metal undercut that define the FinFET device structure. They are critical because they directly impact the electrical performance and characteristics of transistors. We used CDSEM and OCD tools while making reference measurements using a transmission electron microscope (TEM), and compared data from the proposed hybrid OCD (with OCD as the primary tool and CDSEM as the secondary tool) with data from conventional OCD.
We verified the improved accuracy of hybrid metrology over conventional metrology using total measurement uncertainty (TMU) analysis.5, 6 TMU, a measure of variance that represents scatter around a best-fit line, is a useful tool for determining how well a tool under test correlates to an appropriate reference measurement system. We computed the TMU metrics as well as the correlation R2(a statistical measure of how well a regression line approximates real data points). We determined the value of these metrics for metal CD and metal undercut for both hybrid metrology (hybrid OCD) and conventional metrology (conventional OCD). For both parameters, we found hybrid OCD gave significant improvements in normalized TMU, R2 and slope: see Figure 2(b). Further investigation shown that conventional OCD suffered from correlation among fin parameters such as fin CD, fin oxide thickness and fin height. This issue was resolved in hybrid metrology through use of fin CD data from CDSEM. Other modes of hybrid metrology, such as ‘hybrid CDSEM’ (where CDSEM is used as the primary tool and OCD is used as the secondary tool) as well as ‘OCD assisting CDAFM’ have also shown improved measurement performance over conventional individual techniques.1
Advanced hybrid metrology setup where all information is used concurrently for unified optimization.1
CDSEM: CD scanning electron microcopy. CDAFM: CD atomic force microscopy.
A ‘holistic metrology’ approach consists of combining all available information from various sources. This may include hybrid metrology,1 multiple structures,7 and optical channels8to provide an optimum metrology solution with improved measurement performance. A more general or holistic hybrid metrology approach would be for multiple types of chosen toolsets (such as tool A, B, C, and so on) to provide raw data (spectra, a trace, or an image) to a ‘universal hybrid engine’ server that processed the information concurrently (see Figure 3). This mode, though technically the most complex, should provide the maximum benefit in measurement performance. At the same time, any hybrid metrology setup must take measurement time and cost into account.
In summary, the hybrid metrology approach investigated here yielded more accurate, less noisy and less process-sensitive results compared with conventional individual techniques. These traits are necessary for any measurement solution, and especially for advanced 3D structures where there is currently a gap in metrology capabilities. However, to attain the benefits of hybrid metrology, industry (tool suppliers, chip manufacturers, research consortia) will need to collaborate extensively and invest in algorithm development, as well as standardization of protocols and software infrastructures. Future work will focus on designing a universal hybrid engine and optimizing the measurement time and costs.
This work was the effort of a team of 20 scientists and engineers spread across GLOBALFOUNDRIES, IBM and Nova. This work has been supported by independent silicon-on-insulator technology development projects at the IBM Microelectronics Division, Semiconductor Research and Development Center, Hopewell Junction, NY.
Hopewell Junction, NY
Alok Vaid is currently working in the area of metrology research and development. He has published several scientific papers dealing with metrology and process control. He received his Master's degree in Mechanical Engineering from the University of Texas at Austin.
Hopewell Junction, NY
Narender Rana received his PhD in 2007 from the College of Nanoscale Science and Engineering, University at Albany, NY. Since 2007, he has been working as a scientist at the IBM Semiconductor Research and Development Center, Fishkill, NY. He is a metrology expert, involved in international collaboration for metrology development, and leads the reference metrology development at IBM.
Nova Measuring Instruments
Cornel Bozdog received his PhD in Solid State Physics from Lehigh University in 2001. Since 2003, he has been working on the development of optical metrology applications, methodology and analysis algorithms for semiconductor device process control and characterization. He is a Nova Expert fellow currently researching next-generation metrology tools.
1. A. Vaid, B. B. Yan, Y. T. Jiang, M. Kelling, C. Hartig, J. Allgair, P. Ebersbach, A holistic metrology approach: hybrid metrology utilizing scatterometry, CD-AFM, and CD-SEM, Proc. SPIE
7971, pp. 797103, 2011. doi:10.1117/12.881632
2. R. Silver, N. F. Zhang, B. M. Barnes, H. Zhou, A. Heckert, R. Dixson, T. A. Germer, Improving optical measurement accuracy using multi-technique nested uncertainties, Proc. SPIE
7272, pp. 727202, 2009. doi:10.1117/12.816569
3. N. Rana, C. Archie, Hybrid reference metrology exploiting patterning simulation, Proc. SPIE
7638, pp. 76380W, 2010. doi:10.1117/12.848311
4. J. Foucher, N. Rana, 3D-AFM enhancement for CD metrology dedicated to lithography sub-28-nm node requirements, Proc. SPIE
7638, pp. 763802, 2010. doi:10.1117/12.846550
5. M. Sendelbach, C. N. Archie, Scatterometry measurement precision and accuracy below 70nm, Proc. SPIE
5038, pp. 224-238, 2003. doi:10.1117/12.488117
6. M. Sendelbach, A. Munoz, K. A. Bandy, D. Prager, M. Funk, Integrated scatterometry in high-volume manufacturing for polysilicon gate etch control, Proc. SPIE
6152, pp. 61520F, 2006. doi:10.1117/12.659946
7. C. Bozdog, H. Kim, S. Emans, B. Sherman, I. Turovets, R. Urensky, B. Brill, A holistic metrology approach: multi-channel scatterometry for complex applications, Proc. SPIE
7971, pp. 797137, 2011. doi:10.1117/12.881638
8. A. Vaid, M. Sendelbach, D. Moore, T. Brunner, N. Felix, P. Rawat, C. Bozdog, Simultaneous measurement of optical properties and geometry of resist using multiple scatterometry targets,J. Micro/Nanolith. MEMS MOEMS 9, no. 0413062010.