Abstract
We obtain the universal evaluations and expressions of measuring uncertainty for all types of rotating-element spectroscopic ellipsometers. We introduce a general data-reduction process to represent the universal analytic functions of the combined standard uncertainties of the ellipsometric sample parameters. To solve the incompleteness of the analytic expressions, we formulate the estimated covariance for the Fourier coefficient means extracted from the radiant flux waveform using a new Fourier analysis. Our approach can be used for optimization of measurement conditions, instrumentation, simulation, standardization, laboratory accreditation, and metrology.
© 2015 Optical Society of America
1. Introduction
One of the main issues in ellipsometry has been always the improvement of uncertainty measurement for spectroscopic ellipsometers (SEs), which was discussed at the 3rd international conference on spectroscopic ellipsometry (ICSE) in 2003 [1]. It is also predicted that this issue will continue to be focused on in the future [2,3]. It is well known that the limits of optics-based measurements are primarily due to not only the signal-to-noise ratio of the measurement system but also the types of measurements being made [4]. In ellipsometry, the latter may be related with the optical configuration, data acquisition method, data-reduction process, type of ellipsometric sample parameters, and measurement setting parameters, where the measurement setting parameters of the azimuthal angles of fixed elements, wavelengths, angles of incidence, etc. have generally user optional attribute [5–9]. Changing the type of the sample under study into the other may influence the measuring uncertainty, since the uncertainty is strongly related with the optical properties of the sample. Thus, not only the SE developers, but also the SE users need a universal method for evaluating and expressing uncertainty of the result of a measurement.
If there is a universal method of evaluating and expressing the measurement uncertainty for SEs, one can choose the type of SE suitable for a sample, optimize the measurement settings, and possibly predict the practical limit for obtaining information of a sample from measurements. In particular, the method might be helpful to not only develop the Mueller-matrix SEs, but also find their optimum measurement conditions since the Mueller-matrix SEs have a huge number of influence quantities, which contribute to the uncertainty of the measurement, compared with the common rotating-element SEs [10–18]. The method can also be used for instrumentation, simulation, standardization, laboratory accreditation, and metrology. Many of the papers presented at the 6th ICSE in 2014 dealt with complex materials that require a number of nontraditional parameters to describe them, increasing the need for measuring uncertainty [2]. However, from a study of earlier works [19–34], there appears to be no generally accepted method for characterizing SE measuring uncertainty [35].
In this work, we focus on measuring uncertainty for multichannel rotating-element spectroscopic ellipsometers (RE-SEs) because various types of multichannel RE-SEs using an integrating photodetector have been developed and widely used in real-time diagnostics [2,3]. We first introduce a general theory for the data-reduction process to derive the universal theoretical expressions for the measuring uncertainties of the ellipsometric sample parameters. To solve the incompleteness of the uncertainty expressions, we derive analytic functions for quantifying the estimated covariance between the Fourier coefficient (FC) means extracted from a detected radiant flux waveform using a new Fourier analysis algorithm [36]. Finally, we compare the data obtained with the functions with the experimental data.
2. Universal expressions of measuring uncertainty
The universal method for evaluating and expressing the uncertainty of the result of a measurement should be applicable to all kinds of measurements and all types of input data used in measurements. In ellipsometry, the radiant fluxes or exposures measured by a photodetector element are determined on the basis of a series of observations. Then, the measured data are transformed into the ellipsometric sample parameters using the mathematical model, i.e., data-reduction process. Therefore, the universal data-reduction process is of critical importance because, in addition to uncertainty in the observations, it generally includes various influence quantities that are inexactly known. This lack of knowledge contributes to the uncertainty of the measurement, as do the variations of the repeated observations and any uncertainty associated with the mathematical model itself.
In general RE-SE ellipsometric configurations [5–9,37–41], the collimated light beam from the light source passes through a polarization state generator (PSG), is reflected from (or transmitted through) the sample, and passes through the polarization state analyzer (PSA) to impinge the photodetector element (PDE), which transforms the detected radiant flux into electrical signals. The rotatable elements composed of the linearly polarizing and compensating elements are appropriately positioned at the PSG and PSA depending on the optical configurations of the RE-SE type. At least one of the rotatable elements must rotate with a constant angular frequency with others fixed at selected azimuthal angles. For proper measurements, the azimuthal angle positions of the characteristic axes of the rotatable elements should be measured from a real origin defined by an axis that is parallel to the plane of incidence and normal to the sample surface. These azimuth angles are measured from the real origin using the right-hand rule: the thumb is pointed in the direction of light propagation and the other fingers then point toward increasing angle. For brevity, the elements are assumed to be ideal while the angle of incidence , wavelength , and other system parameters are known to arbitrary uncertainty.
In the Stokes representation [42,43], the Stokes vectors of and can be described for a light wave incident along the path through the PSG and PDE, respectively, and the Mueller matrices of , , , and for the sample, the PSG, the PSA, and the detector optic system (deployed between the PSA and PDE), respectively. Here, and denote PSG and PSA effective transmittances, respectively, and the superscript is the transpose of a matrix. The Stokes vector of the monochromatic light wave incident on a PDE is described as
The radiant flux waveform detected by a PDE with active area and quantum efficiency is:where denotes the azimuthal angle of the constantly rotating element with respect to the real origin, the dc FC, and and the ac FCs. Using a column vector with the entries of , where the integers and depend on the RE-SE type used, the FCs can be represented as the scalar product given by where is a common factor, , , , , , and . If an RE-SE has a part of or an entire of a fixed polarizer as the first element in the PSG and a fixed analyzer as the last element in the PSA along the light wave train starting from the light source, then is given aswhere , , , and the azimuthal angles and of the fixed analyzer and the fixed polarizer, respectively, are measured with respect to the real origin. Therefore, if the value of the common factor is determined for a RE-SE used, the values of the FCs for a given sample can be calculated theoretically using Eqs. (3)–(5).Upon setting and , Eqs. (3)-(5) can be expressed as . If is nonsingular, the sample vector can be obtained as
where the matrix of the sensitivity coefficients is . Thus, each entry of the sample vector of Eq. (7) is not measured directly, but determined from other quantities through the functional relationship of , where is obtained by a series of observations. is also a function of different uncertainty sources (not only the repeated observations) described as of which the variables may be the wavelength of the light detected in a PDE, the angle of incidence, the azimuthal angles of the fixed elements, and so on. Then, the combined standard uncertainty of for a sample is given by [44]where is the estimated covariance associated with the FC means, is the standard uncertainty of , the partial derivatives are evaluated at , and the over-bar denotes the arithmetic mean. The first and second terms on the right-hand side in Eq. (8) may be called the type A and B evaluations of standard uncertainty, respectively [44].For isotropic samples with a uniform reflecting surface, , , and have been used as the sample parameters [42,43]. Thus, if a sample parameter is described as a function of the entries of , its combined standard uncertainty of is given by [44]
Therefore, one may evaluate the uncertainty of an RE-SE theoretically using Eq. (8) or Eq. (9). However, the theoretical expressions require the analytic functions for the estimated covariance between the FC means, which have not previously been calculated [45].3. Modeling of covariance between Fourier coefficients
To obtain the theoretical expressions for the covariance between the FC means, we want more detailed analysis of the data acquisition method for the RE-SEs. For measurement using an RE-SE, the output signals of the radiant flux values of the quasi-monochromatic light collected by the PDE at time can be expressed using Eq. (2) as:
where denotes the experimental dc FC, and the experimental ac FCs, the angular frequency of the constantly rotating element with revolution period , and the fluctuations due to random noises. In state-of-the-art real-time RE-SEs, each pixel or each binning pixel group in the CCD or photodiode arrays acts as a PDE. When the PDE is read out times during the interval with equal scanning times of , the measured exposure data for an arbitrarily given integration time are generated as , where denotes the time delay prior to beginning the integration after receiving a trigger pulse (one of the clock pulses equidistantly generated from the optical encoder in the rotational stage for the constantly rotating element). The measured exposures can be expressed aswhere , , and . A full set of exposures can be represented as , where , , , , and . If is nonsingular the solutions can be obtained by the least-squares estimation given asTo obtain the solutions to the abovementioned equations, we recently introduced a concise theory using a new method based on the discrete Fourier transform [36], which is applied to the measured as
where and denote real functions, and the over-bar denotes the arithmetic mean of independent pairs of repeated observations. Using the orthogonality of the system of trigonometric functions, the sample mean values of the FCs can be obtained as where , , and . The two data acquisition methods of the least-squares estimation (Eq. (12)) and the discrete Fourier transform on the exposures (Eqs. (14)-(16)) can gives the same values of .To obtain the theoretical expression of the covariance between the FC means, we will adopt the latter data acquisition method. We assume that the random noise in Eqs. (10) and (11) originates from photon noise. Photon noise follows the Poisson distribution, wherein the population variance of the number of the photons detected by a PDE is equal to the arithmetic mean of the same [46]. Therefore, under the photon-noise-limited condition, the population variance of the randomly-varying exposure is given as
where the scaling factor has the same dimension as , and denotes the average photon energy of the quasi-monochromatic photons incident on the PDE. Since the exposures measured in one integration time are not correlated with the other exposures, the estimate of the covariance associated with and is given bywhere denotes the Kronecker delta. Therefore, using the relation of Eq. (18), we obtain the estimate covariance associated with the measured FC means of Eqs. (14)-(16) as wherein and the symmetrical relations of and , obtained by Eqs. (15)-(16), should be used for the special cases of . In the limit of , the autocorrelations of Eqs. (19)-(21) converge to the earlier results [27] for the non-integrating PDEs. For the FCs of , the estimated standard uncertainties and the estimated correlation coefficients are given asFor proper measurements, the azimuthal angle positions of the characteristic axes of the rotatable elements should be corrected for the real origin using the calibration procedures [5–9]. If the azimuthal angle of the constantly rotating element is represented as with respect to the real origin, in which denotes the th order offset angle, the relationship between the original (unprimed) and experimental (primed) FCs can be obtained from the equality relation between Eqs. (2) and (10) as
from which the observed values of the original FCs can be obtained. If the uncertainty of due to random noises in can be ignored for state-of-the-art RE-SEs, we obtain . Therefore, if the value of the scaling factor for each PDE in a RE-SE used is known, the estimated covariance in Eq. (8) and (9) can be determined quantitatively from the theoretical FC functions of Eqs. (3)-(5) through Eqs. (19)-(26).4. Application
To examine the relationship of the reliability of the derived stochastic model functions and the estimated covariance between the FC means, we adopt a bare Si wafer as a sample and a homemade multichannel rotating-polarizer SE based on a three-polarizer design [45]. In our trials, each spectrum of the original FCs for the sample was measured at multiple azimuthal angles of the fixed analyzer from 0° to 358° in steps of 2° for the incidence angle of 70°; the measured data are shown in Fig. 1.
The observed values of the sample standard uncertainties and the sample correlation coefficients for the original FCs in Fig. 1 can be given as [44]
of which the experimental data are shown as circles in Figs. 2 and 3, respectively. In the study, using the relation of Eq. (19) and the corresponding experimental data, the mean of the data set of , which was calculated at each scanning angle of the fixed analyzer, was obtained as count for 550 nm. Since 18.23°, in the third case of Eq. (6) is constant. In Eq. (3), the values of can be calculated by using the known optical properties of the sample, such as the refractive index of c-Si (4.077 - i0.025) and the refractive index (1.465) and film thickness (1.97 nm) of the native oxide, and the values of can be determined by the known and values. Thus, the unknown parameters in Eq. (6) were determined as count/ms, , and by using the least-squares fitting for the function of Eq. (3) and the corresponding experimental values. Substituting the calculated FCs into Eqs. (27) and (28), we obtained theoretically the estimated standard uncertainty and estimated correlation coefficient values, which also closely agreed with the corresponding experimental data, as shown in Figs. 2 and 3. We remark here that and can be derived as the functions of the five FCs {, , , , } using the universal data-reduction method, and subsequently, their partial derivative functions can be obtained from the derivatives of the sample parameter functions with respect to the entries {, , } of the sample Mueller matrix. Consequently, by substituting the partial derivative, sensitivity coefficient, and covariance functions into the type A evaluation term in Eq. (9), we obtained the theoretical values of the combined standard uncertainties, and , of and for the repeated observations shown in Fig. 4. These analytic expressions exhibit excellent agreement with their corresponding experimental data (Fig. 4). The azimuthal angle dependency for and, as shown in Fig. 4, might be changed depending on both the wavelength and the type of sample parameters, i.e., their minimum positions are different in general. Therefore, to determine the optimized conditions for spectroscopic data, we define an uncertainty test function as follows:where is the number of types of sample parameters and denotes the number of measuring wavelengths. The analytic expression of can be obtained as a function of the azimuthal angles of the fixed elements and the angle of incidence, and subsequently used to find the optimized conditions in which the value of is minimum. In conclusion, we believe our approach can significantly aid in evaluating SE uncertainty in all types of multichannel RE-SEs.5. Conclusion
In this work, for the first time, we obtained universal analytic expressions of the measuring uncertainty for multichannel RE-SEs. In the limit of , the analytic expressions converge to the measuring uncertainties of the RE-SEs adopting the non-integrating PDEs. A general theory of data-reduction procedures was introduced to obtain universal expressions of the theoretical combined standard uncertainties of the ellipsometric sample parameters. This required the determination of the analytic functions for the estimated covariance between the FC means, which has been undetermined thus far. Using a new closed-form Fourier analysis, we derived the analytic functions for estimated covariance between the FC means extracted from the detected radiant flux waveform. Since the uncertainties of the ellipsometric sample parameters are strongly related with not only the type of the sample parameters but also the wavelength, we introduced the uncertainty test function of Eq. (34) to find the optimal measurement conditions of an RE-SE for a given sample. The analytic expressions can be used for instrumentation, simulation, standardization, laboratory accreditation, and metrology. This approach will be also helpful for evaluating and expressing the measuring uncertainties for other types of SEs, such as photoelastic modulator SEs [47,48].
Acknowledgments
This work was supported by the Korea Research Institute of Standards and Science under the project “Development of core measurement technologies for the next generation of nano devices” (Grant No. 15011043).
References and links
1. D. E. Aspnes, “Expanding horizons: new developments in ellipsometry and polarimetry,” in Proceeding of the 3rd International Conference on Spectroscopic Ellipsometry, M. Fried, K. Hingerl, and J. Humlicek, ed. Thin Solid Films 455–456, 3–13 (2004). [CrossRef]
2. D. E. Aspnes, ” Spectroscopic ellipsometry – Past, present, and future,” in Proceeding of the 6th International Conference on Spectroscopic Ellipsometry, Y. Otani, M. Tazawa, and S. Kawabata, ed. Thin Solid Films 571, 334–344 (2014).
3. S. Zollner, “Spectroscopic Ellipsometry for Inline Process Control in the Semiconductor Industry,” in Ellipsometry at the Nanoscale, M. Losurdo and K. Hingerl, ed. (Springer, Berlin, 2013), Chap. 18.
4. J. Opsal, “How far can one go with optical metrology?” Laser Focus World 42, 101–105 (2006).
5. R. M. A. Azzam and N. M. Bashara, Ellipsometry and Polarized Light (North-Holland, Amsterdam, 1987).
6. H. G. Tompkins and E. A. Irene, Handbook of Ellipsometry (Williams Andrew Inc., 2005).
7. H. Fujiwara, Spectroscopic ellipsometry: Principles and Applications (John Wiley & Sons, Ltd., 2007).
8. M. Schubert, Infrared Ellipsometry on Semiconductor Layer Structures: Phonons, Plasmons, and Polaritons (Springer, Berlin, 2010).
9. M. Losurdo and K. Hingerl, Ellipsometry at the Nanoscale (Springer, Berlin, 2013).
10. R. M. A. Azzam, “Photopolarimetric measurement of the Mueller matrix by Fourier analysis of a single detected signal,” Opt. Lett. 2(6), 148–150 (1978). [CrossRef] [PubMed]
11. P. S. Hauge, “Mueller matrix ellipsometry with imperfect compensators,” J. Opt. Soc. Am. 68(11), 1519–1528 (1978). [CrossRef]
12. D. H. Goldstein, “Mueller matrix dual-rotating retarder polarimeter,” Appl. Opt. 31(31), 6676–6683 (1992). [CrossRef] [PubMed]
13. J. Lee, J. Koh, and R. W. Collins, “Dual rotating-compensator multichannel ellipsometer: Instrument development for high-speed Mueller matrix spectroscopy of surfaces and thin films,” Rev. Sci. Instrum. 72(3), 1742–1754 (2001). [CrossRef]
14. M. H. Smith, “Optimization of a dual-rotating-retarder Mueller matrix polarimeter,” Appl. Opt. 41(13), 2488–2493 (2002). [CrossRef] [PubMed]
15. C. Chen, I. An, and R. W. Collins, “Multichannel Mueller Matrix Ellipsometry for Simultaneous Real-Time Measurement of Bulk Isotropic and Surface Anisotropic Complex Dielectric Functions of Semiconductors,” Phys. Rev. Lett. 90(21), 217402 (2003). [CrossRef] [PubMed]
16. L. Broch, A. En Naciri, and L. Johann, “Second-order systematic errors in Mueller matrix dual rotating compensator ellipsometry,” Appl. Opt. 49(17), 3250–3258 (2010). [CrossRef] [PubMed]
17. J. Li, B. Ramanujam, and R. W. Collins, “Dual rotating compensator ellipsometry: Theory and simulations,” Thin Solid Films 519(9), 2725–2729 (2011). [CrossRef]
18. E. Garcia-Caurel, R. Ossikovski, M. Foldyna, A. Pierangelo, B. Drévillon, and A. D. Martino, “Advanced Mueller Ellipsometry Instrumentation and Data Analysis,” in Ellipsometry at the Nanoscale, M. Losurdo and K. Hingerl, ed. (Springer, Berlin, 2013), Chap. 2.
19. E. Schmidt, “Precision of Ellipsometric Measurement,” J. Opt. Soc. Am. 60(4), 490–494 (1970). [CrossRef]
20. D. E. Aspnes, “Precision Bounds to Ellipsometer Systems,” Appl. Opt. 14(5), 1131–1136 (1975). [CrossRef] [PubMed]
21. J. M. M. de Nijs and A. V. Silfhout, “Systematic and random errors in rotating-analyzer ellipsometry,” J. Opt. Soc. Am. A 5(6), 773–781 (1988). [CrossRef]
22. D. H. Goldstein and R. A. Chipman, “Error analysis of a Mueller matrix polarimeter,” J. Opt. Soc. Am. A 7(4), 693–700 (1990). [CrossRef]
23. N. V. Nguyen, B. S. Pudliner, I. An, and R. W. Collins, “Error correction for calibration and data reduction in rotating-polarizer ellipsometry: applications to a novel multichannel ellipsometer,” J. Opt. Soc. Am. A 8(6), 919–931 (1991). [CrossRef]
24. I. An and R. W. Collins, “Waveform analysis with optical multichannel detectors: Applications for rapid-scan spectroscopic ellipsometry,” Rev. Sci. Instrum. 62(8), 1904–1911 (1991). [CrossRef]
25. R. Kleim, L. Kuntzler, and A. El Ghemmaz, “Systematic errors in rotating-compensator ellipsometry,” J. Opt. Soc. Am. A 11(9), 2550–2559 (1994). [CrossRef]
26. Z. Huang and J. Chu, “Optimizing precision of fixed-polarizer, rotating-polarizer, sample, and fixed-analyzer spectroscopic ellipsometry,” Appl. Opt. 39(34), 6390–6395 (2000). [CrossRef] [PubMed]
27. D. E. Aspnes, “Optimizing precision of rotating-analyzer and rotating-compensator ellipsometers,” J. Opt. Soc. Am. A 21(3), 403–410 (2004). [CrossRef] [PubMed]
28. B. Johs and C. M. Herzinger, “Precision in ellipsometrically determined sample parameters: simulation and experiment,” Thin Solid Films 455–456, 66–71 (2004). [CrossRef]
29. J. A. Zapien, A. S. Ferlauto, and R. W. Collins, “Application of spectral and temporal weighted error functions for data analysis in real-time spectroscopic ellipsometry,” Thin Solid Films 455–456, 106–111 (2004). [CrossRef]
30. K. M. Twietmeyer and R. A. Chipman, “Optimization of Mueller matrix polarimeters in the presence of error sources,” Opt. Express 16(15), 11589–11603 (2008). [CrossRef] [PubMed]
31. L. Broch, A. En Naciri, and L. Johann, “Systematic errors for a Mueller matrix dual rotating compensator ellipsometer,” Opt. Express 16(12), 8814–8824 (2008). [CrossRef] [PubMed]
32. B. Johs and C. M. Herzinger, “Quantifying the accuracy of ellipsometer systems,” Phys. Status Solidi C 5(5), 1031–1035 (2008). [CrossRef]
33. L. Broch and L. Johann, “Optimizing precision of rotating compensator ellipsometry,” Phys. Status Solidi C 5(5), 1036–1040 (2008). [CrossRef]
34. L. Broch, A. E. Naciri, and L. Johann, “Analysis of systematic errors in Mueller matrix ellipsometry as a function of the retardance of the dual rotating compensators,” Thin Solid Films 519(9), 2601–2603 (2011). [CrossRef]
35. M. Losurdo, “Applications of ellipsometry in nanoscale science: Needs, status, achievements and future challenges,” in Proceeding of the 5th International Conference on Spectroscopic Ellipsometry, edited by H. G. Tompkins, Thin Solid Films 519, 2575–2583 (2011). [CrossRef]
36. Y. J. Cho, W. Chegal, and H. M. Cho, “Fourier analysis for rotating-element ellipsometers,” Opt. Lett. 36(2), 118–120 (2011). [CrossRef] [PubMed]
37. P. S. Hauge, “Recent development in instrumentation in ellipsometry,” Surf. Sci. 96(1-3), 108–140 (1980). [CrossRef]
38. R. W. Collins, “Automatic rotating element ellipsometers: Calibration, operation, and real-time applications,” Rev. Sci. Instrum. 61(8), 2029–2062 (1990). [CrossRef]
39. K. Vedam, “Spectroscopic ellipsometry: a historical overview,” Thin Solid Films 313–314, 1–9 (1998). [CrossRef]
40. R. M. A. Azzam, “The intertwined history of polarimetry and ellipsometry,” Thin Solid Films 519(9), 2584–2588 (2011). [CrossRef]
41. D. E. Aspnes, “Spectroscopic ellipsometry – A perspective,” J. Vac. Sci. Technol. A 31(5), 058502 (2013). [CrossRef]
42. G. E. Jellison, Jr., “Data Analysis for Spectroscopic Ellipsometry,” in Handbook of Ellipsometry, H. G. Tompkins and E. A. Irene, ed. (Williams Andrew Inc., 2005), Chap. 3.
43. D. Schmidt, E. Schubert, and M. Schubert, “Generalized Ellipsometry Characterization of Sculptured Thin Films Made by Glancing Angle Deposition,” in Ellipsometry at the Nanoscale, M. Losurdo and K. Hingerl, ed. (Springer, Berlin, 2013), Chap. 10.
44. BIPM, IEC, IFCC, ISO, IUPAC, IUPAP, and OIML, JCGM 100:2008, “Evaluation of measurement data - Guide to the expression of uncertainty in measurement,” (GUM 1995 with minor corrections) (www.bipm.org).
45. W. Chegal, J. P. Lee, H. M. Cho, S. W. Han, and Y. J. Cho, “Optimizing the precision of a multichannel three-polarizer spectroscopic ellipsometer,” J. Opt. Soc. Am. A 30(7), 1310–1319 (2013). [CrossRef] [PubMed]
46. D. J. Brady, Optical Imaging and Spectroscopy (John Wiley & Sons, Ltd., 2009), Chap. 5.
47. B. Drévillon, “Phase modulated ellipsometry from the ultraviolet to the infrared: In situ application to the growth of semiconductors,” Prog. Crystal Growth and Charact. 27(1), 1–87 (1993). [CrossRef]
48. G. E. Jellison, Jr. and F. A. Modine, “Polarization Modulation Ellipsometry,” in Handbook of Ellipsometry, H. G. Tompkins and E. A. Irene, ed. (Williams Andrew Inc., 2005), Chap. 6.