Abstract
In this Letter, we describe a new, to the best of our knowledge, concept of angle-resolved spectral reflectometry with a pixelated polarizing camera for determination of the thickness of each film layer in multilayer films. We can measure the changes in the phase and amplitude of p- and s-polarized light over a broad spectral range and a wide incident angle at a time. The proposed method is verified by measuring a sample of multilayer film and comparing our measurement results with an ellipsometer. The comparison results show that our proposed technique enables real-time inspection of multilayer films with high precision.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
An ellipsometer [1–3] is widely used for the investigation of multilayer film stacks in the semiconductor industry. It is capable of measuring very important film parameters such as the thickness and refractive index, which can be used as a quality check of the films. However, the ellipsometer is difficult to use as production inspection equipment because of being configured off-axis and low spatial resolution due to the large spot size of 25 µm or greater. Moreover, it takes long measurement times, because it needs many measurement steps in order to measure the changes in the state of polarization of reflected lights by changing the incident angle and wavelength. In response to these continuous industrial demands for high spatial resolution and high speed, many studies have been performed in the film measurement field [3–9]. Among them, the beam profile reflectometer technique [4–6] can be a suitable candidate to fulfill these needs. Its on-axis configuration with a high NA objective provides submicron resolution, and it measures multi-angular reflectance for various polarization states in real time. However, this approach is limited to monochromatic light. In order to obtain multi-angular reflectance over various wavelengths, we need to vary the wavelength of the light source sequentially. For simultaneous measurements of both the spectral and angular variations of reflectance, angle-resolved spectral reflectometry was developed [10]. However, it measures only p-polarized or s-polarized reflectance according to the polarization state of the incident light and the direction of a line slit. In order to overcome all the technical limitations described above, we suggest a new concept of angle-resolved spectral reflectometry. Our proposed method can obtain the absolute reflectance data of the sample over a broad spectral range and a wide incident angle according to various polarization states in a single-shot measurement.
Figure 1 shows a schematic diagram of our proposed single-shot angle-resolved spectral reflectometry. A broad-spectrum light source focuses on the sample through a high numerical aperture objective. The focused light beam consists of many ray bundles with the angle of incidence ranging from $ - {{\rm \sin}^{ - 1}}({\rm NA})^\circ $ to ${{\rm \sin}^{ - 1}}({\rm NA})^\circ $, and each ray strikes the sample surface at its incident angle. Thus, a high NA objective lens is required to measure a wide range of angular reflectance according to the incident angle. The focused incident beam undergoes an infinite loop of multiple reflections within the sample, and it creates the interference fringe pattern. Then the back-reflected beam forms an angular reflectance fringe pattern at the back focal plane of the objective lens, and the radial direction of this circular-fringe interferogram represents the incidence angle. This angular interferogram results from the sum of monochromatic individual interference patterns over a broad spectral range.
Relay optics is used to deliver the interference fringe pattern formed at the back focal plane of the objective into an entry line slit of the imaging spectrometer, and the line slit passes only the selected line of the radial-circular interference pattern, as shown in Fig. 2(a). The relation between the radial pixel position ($r$) and the incident angle ($\theta $) can be expressed as follows:
To verify our proposed method, we fabricated a multi-stacked dielectric film layer patterned by a photolithography process. Figure 3 shows details of the internal structure of the sample. Three types of films with various thicknesses are sequentially stacked on a patterned silicon wafer in the order of ${{\rm Si}_3}{{\rm N}_4} {\text -} {{\rm SiO}_2} {\text -} {\rm SiON} {\text -} {{\rm SiO}_2} {\text -} {{\rm Si}_3}{{\rm N}_4}$. We performed measurement on two spots at T1 and T2 positions, of which the nominal thickness sets are (300, 350, 450, 700, 600 nm) and (300, 550, 350, 500, 1300 nm), respectively.
In this experiment, we used commercial products for an imaging spectrometer (V8E, SPECIM) and a pixelated polarizing camera (BFS-U3-51S5, FLIR), respectively. Table 1 shows details of our optical system used in the experiments. Once four angle-resolved spectral reflectance images corresponding to each sub-array in the pixelated polarizer camera are measured, we can obtain the angle-resolved spectral images of p- and s-polarization reflectance and sine of the phase difference between p- and s-polarized light, respectively, using Eq. (3).
Before measurement, we performed the preliminary work of calibrating both the vertical axis and the horizontal axis of the angle-resolved spectral interferogram. First, we calibrated the wavelength of the vertical axis by finding the relation between the pixel position and wavelength using a mercury argon (HgAr) source. We used three narrow, intense peaks (404.66, 435.84, 546.07 nm) from the source to determine the wavelength calibration equation, a third-order polynomial fitting model for mapping the pixel to wavelength. Next, we calibrated the incident angle of the horizontal axis by mapping the pixel position to the incident angle using Eq. (1). Here ${\theta _{{\rm max}}}$ can be determined by finding the pixel position (${r_B}$) corresponding to Brewster’s angle (${\theta _B} = {56.6}^\circ $) of BK 7 glass with a refractive index of 1.5185 at 550 nm using the relation of ${{\theta }_{\max }}={{\sin }^{-1}}( {r}_{\max }/{r}_{B}\times \sin {{\theta }_{B}}).$ After these calibration works, we performed measurements for the multilayer sample and compared our results with other conventional ellipsometer techniques. Figure 4 shows the comparison results of experiments and simulations of ${\Re _p}$, ${\Re _s}$, and ${\rm \sin}\Delta $, respectively, for the five-layer film sample.
The optimal sets of the film thickness of each layer of the sample at T1 and T2 positions (${d_1}$, ${d_2}$, ${d_3}$, ${d_4}$, ${d_5}$), were numerically determined by minimizing the merit function of Eq. (4) by a global optimization technique. Each theoretical model for p-polarized reflectance, s-polarized reflectance, and phase difference are well matched to the measurements. However, the experimental measurements tend to be less sharp than the simulated data at critical points, because the measured angular spectra data at each pixel with finite size actually is the response containing the contribution of a small range of angles and wavelength, rather than a single angle and wavelength. If reflecting these considerations in the simulations, the simulations will agree better with the measurement results. Table 2 shows the comparisons of our measurement results with a commercial instrument, the KLA-Tencor SpectraFx 100 ellipsometer. The thickness deviation of each layer between the two methods is less than ${\sim}11.3\;{\rm nm}$, and the comparison results show good agreement with each other.
To conclude, we have proposed and verified a new method of single-shot angle-resolved spectral reflectometry by comparing our measurement results with an ellipsometer; our method shows performance equivalent to the conventional methods. Our method acquires only a single image, which simultaneously provides the changes in the phase and amplitude of p-and s-polarized light reflected from the sample. Hence, we can measure the film thickness corresponding to each layer of multilayer film in real time. Moreover, its on-axis configuration provides higher spatial resolution and a relatively compact setup. We expect that our proposed technique will be widely used as an in-line metrology tool, especially in the semiconductor industry in terms of high measurement speed, robustness to external noises, and high spatial resolution.
Funding
Korea Research Institute of Standards and Science (19011086); Ministry of Trade, Industry and Energy (18201022).
Disclosures
The authors declare no conflicts of interest.
REFERENCES
1. R. M. A. Azzam and N. M. Bashra, Ellipsometry and Polarized Light (North-Holland, 1987).
2. H. G. Tompkins and W. A. McGahan, Spectroscopic Ellipsometry and Reflectometry: A User’s Guide (Wiley, 1999).
3. T. E. Jenkins, J. Phys. D 32, R45 (1999). [CrossRef]
4. A. Rosencwaig, J. Opsal, D. L. Willenborg, S. M. Kelso, and J. T. Fanton, Appl. Phys. Lett. 60, 1301 (1992). [CrossRef]
5. J. T. Fanton, J. Opsal, A. Rosencwaig, and D. L. Willenborg, Proc. SPIE 2004, 313 (1993). [CrossRef]
6. J. M. Leng, J. Opsal, and D. E. Asppnes, J. Vac. Sci. Technol. A 17, 380 (1999). [CrossRef]
7. K. Wu, C.-C. Lee, N. J. Brock, and B. Kimbrough, Opt. Lett. 36, 3269 (2011). [CrossRef]
8. D. Kim, Y. Yun, and K.-N. Joo, Opt. Lett. 42, 3189 (2017). [CrossRef]
9. Y.-S. Ghim, H.-G. Rhee, and A. Davies, Sci. Rep. 7, 11843 (2017). [CrossRef]
10. J. Dong and R. Lu, Opt. Express 26, 12291 (2018). [CrossRef]
11. G. E. Jellison Jr., Appl. Opt. 30, 3354 (1991). [CrossRef]