High spectral purity at longer wavelength side is demanded in many extreme ultraviolet (EUV) and soft X-ray (together also referred to as XUV) optical systems. It is usually obtained at the expense of a high loss of XUV efficiency. We proposed and developed a new method based on a periodic, tapered structure integrated with an EUV multilayer. The longer wavelength radiation is scattered/diffracted away by the tapered multilayer structure while the EUV light is reflected. The first proof-of-principle showed a broadband suppression from λ = 100-400 nm with an average factor of 14. Moreover, a high EUV reflectance of 64.7% was achieved, which corresponds to 94% of the efficiency of a regular EUV multilayer mirror.
© 2014 Optical Society of America
Development of advanced XUV (Extreme ultraviolet and soft X-ray) applications and their sources have brought up new requirements on the spectral characteristics of the optics, like the efficiency, bandwidth, spectral purity and so on. Spectral contamination at the longer wavelength range is a common problem of many XUV sources. Plasma sources, both laser and discharge produced, have a wide emission spectrum from the EUV to visible, sometimes extending into the infrared range [1,2]. High harmonic generation sources produce a harmonic train of wavelengths starting from the principal wavelength of the drive laser . In astronomy observations targeting at the EUV region, a much brighter background of longer wavelengths will also be present in the spectrum . The multilayer (ML) mirror is one of the essential optics working in the XUV region . It enables the reflection of these short wavelengths at non-grazing angles of incidence. However, the longer wavelength light in the spectrum will also be reflected by the multilayer and enter the optical system where it can deteriorate the detection sensitivity/imaging resolution  or cause a heat load problem . EUV photolithography is one of the notable examples potentially suffering from this longer wavelength contamination of the spectrum. The EUV plasma sources not only produce the 13.5nm in-band radiation but also a large amount of UV out-of-band light (λ = 100-400nm and beyond). The UV radiation is efficiently reflected by the typically used Mo/Si multilayer coated reflective optics and it can reach the wafer causing imaging contrast loss . A transmission filter can be employed to block the unwanted longer wavelength light. But such filters have a relatively low transmittance of EUV light and are prone to break under heat load or pressure variation . Multilayer mirrors combined with an antireflection layer [10,11] or phase shift grating  have been studied but still have limitations on the suppressed bandwidth. Different blazed gratings were also proposed as a spectral purity filter while the EUV diffraction efficiency needs to be further improved [13,14]. A novel suppression method based on surface pyramids structure has been developed by our group recently . Full band suppression of the UV light (100-400nm) combined with an EUV reflectance of 56.2% has been demonstrated by producing silicon pyramids on top of the multilayer. In this paper, we will show a new scheme of this method by using a multilayer pyramids structure which shows a significant full band suppression effect and a much higher EUV reflectance than the pyramid structure made of silicon.
2. Design of the multilayer pyramid structure
The method of the surface pyramids structure  is based on the fact that EUV light can only be reflected by a periodic multilayer while UV can be reflected by a single layer (due to the relative large contrast of the refractive index in the UV region). Thus, a surface tapered structure, e.g. pyramids, made from an EUV transmitting material (silicon) can scatter/diffract the UV light out of the specular direction, while the EUV is reflected by the ML underneath (typically composed by a Mo/Si bi-layered system for λ = 13.5nm). Since the absorption coefficient of all materials, including silicon, in the EUV region cannot be neglected, a certain amount of EUV power will be lost due to absorption in the pyramids. However, if the pyramid consists of a multilayer, as both Mo and Si are reflective for the UV light, it will have similar UV light suppression effects but the EUV light is now reflected by the ML pyramids and the ML underneath. In this case, no EUV power will be lost in principle. The Si pyramids structure developed in  is tapered in two dimensions (2D) instead of one dimension (1D) to limit the absorption. For ML pyramids, both 2D and 1D structures can be used as shown in Fig. 1.In this paper, we use Mo/Si multilayers with a d-spacing of 7nm to make the pyramids structure.
2.1 UV suppression dependence on height and top-flat area
The UV suppression is mainly caused by “blazed” diffraction from the periodic facets and the interference between the reflection from the top-flat area and the valley in between the pyramids, so that most of the UV power is diffracted to higher orders. This can be optimized by tuning the structural parameters including the height (h), period (p), bottom width (a) and top-flat width (w). The effects of different structural parameters and a detailed optimization for the 2D structure have been discussed in the paper describing Si pyramids . ML pyramids have almost the same UV response as Si pyramids (with the same tapered structure). This also indicates that the UV suppression effect is mainly determined by the geometrical shape while the layers and interfaces in the multilayer pyramids do not change the overall result. The dependence of UV suppression on structural parameters for 1D and 2D pyramids was further studied. Diffraction efficiency of 1D and 2D ML pyramids structure was calculated by using the DiffractMOD of Rsoft software  based on the Rigorous Coupled Wave Analysis . According to the optimization results of the Si pyramids and for easy implementation in the experiments, the ML pyramids heights are both set at h = 105nm, corresponding to 15 bilayers of Mo/Si. The period is equal to the bottom width, p = a = 26µm. Two top-flat area widths of w = 0.06p and 0.34p are used in the calculations. A plane wave at normal incidence is used with linear polarization at the p state. For 1D pyramids, this means that the electric field is oscillating in the plane defined by the ML surface normal and the grating vector shown in Fig. 1. Both the 0th order and the total diffraction efficiency (sum of the efficiency of all diffraction orders including the 0th order) are calculated, as shown in Fig. 2.For pyramids almost without a top-flat (w = 0.06p), both the total diffraction efficiency and the 0th order efficiency of the 2D structure are higher than that of the 1D structure (the 0th order efficiency of 2D and 1D structure is indicated as (0,0) and 0th in Fig. 2, respectively). The reason can be attributed to the different geometrical shape. The effective valley area of 2D pyramids where a large amount of UV power can transmit through the facets and reflected by the plane multilayer underneath is doubled compared to 1D. The energy distribution over various diffraction orders is also different for the two structures. However, for pyramids with a larger top flat area (w = 0.34p), the suppression achieved with the 2D structure is greatly enhanced by the destructive interference between reflection from the top and the valley. The UV reflectance is reduced to below 11% over the entire range of λ = 100-400nm with an average value of only 3.9%. A small bump of reflectance appears around λ = 180nm due to a constructive interference effect. For 1D pyramids with w = 0.34p, the interference effects become much stronger resulting in an even lower reflectance at longer wavelength and a larger bump at shorter wavelength. It means that the reflection from the top and valley of the 1D structure are better matched for interference. Therefore, a much smaller top width is required to avoid the large bump at shorter wavelength. With the help of tuning the top-flat area, 2D pyramids can provide almost the same suppression effect as 1D (with the same height). A larger height can certainly be used for both 2D and 1D structure to increase the blazed angle of the facets and achieve an even higher suppression result. Unlike the Si pyramids, this will in principle not cause any loss of EUV reflectance. 2D ML pyramids with h = 154nm (containing 22 bilayers) and 0.34p top width, or 1D pyramids with h = 154nm and 0.04p top width can reduce the UV reflectance to below 4% over the entire range with an average value of 1.7% and 1.1%, respectively. For the clarity of the graph, the reflectance of larger height pyramids is not shown in Fig. 2.
2.2 UV suppression dependence on the period
The period of the ML pyramids (both 1D and 2D) is not a sensitive factor for UV suppression. The reflectance of 2D pyramids with p = 1µm, 26µm, 50µm, 100 µm was calculated at normal incidence for p-polarized light. It can be seen from Fig. 3 that both the 0th order and the total diffraction efficiency are almost the same for different periods . The results on the1D pyramids have the same trend which is not shown here. The fact that the period of the pyramids has little influence on the UV suppression can be used as an advantage of the method. As the ML pyramids can also diffract the EUV light, it will distribute the power into several diffraction orders. Although the neighboring diffraction orders of the EUV light are very close to the 0th order reflection, they may still cause an illumination problem in lithography and other imaging applications. To avoid this, the structural period can be increased to the value beyond the spatial coherence length of the EUV light to avoid EUV diffraction. Let’s take the laser produced Sn-plasma source  as an example. The coherent length of the emitted 13.5nm light can be roughly estimated as, where z is the distance from the source to the optics, d is the diameter (FWHM) of the source assuming a Gaussian intensity distribution . For the case of z = 20cm, d = 100µm, the spatial coherence length of the EUV light is l13.5nm = 13.5µm, lUV = 100-400µm. Thus, if the period of the pyramids is selected as l13.5nm<p<lUV, a full band UV suppression can be achieved while no EUV diffraction orders will be generated.
2.3 UV suppression dependence on polarization and incidence angle
Besides the structural parameters, the incidence angle and polarization of the light will also affect the UV suppression effect, especially for the 1D ML pyramids. Two incident angles need to be considered, the in-plane angle (φ) and out-of-plane angle (θ), as shown in Fig. 1. This will be discussed together with the polarization effect. For 2D pyramids, the suppression effect at normal incidence is not sensitive to polarization since the pyramid is a symmetric structure in two dimensions. The orientation of the E-vector will only affect the efficiency of high diffraction orders in space, but the 0th order and total diffraction efficiency are the same. However, the suppression effect of 1D pyramids is more sensitive to polarization as is illustrated in Fig. 4.To show the potential of UV suppression, the 1D pyramids structure calculated here is h = 154nm (22 bilayers), w = 0.04p, p = a = 26µm. At normal incidence with φ = 0°, θ = 0°, the total diffraction efficiency for s-polarized light is much higher than for p-polarization, while the 0th order efficiency is only slightly higher (p-polarization has been defined above). As the incidence geometry changes to φ = 0°, θ = 90°, the orientation of the E-vector with respect to the grating vector is reversed for s- and p-polarization. The curves for p-polarized light are then exactly the same as in the s-polarized case of φ = 0°, θ = 0° (not shown in the graph). This means that the dependence of UV suppression of 1D pyramids on polarization (at normal incidence) is mainly affected by the orientation of the E-vector with respect to the grating vector (the modulation direction of the refractive index) . If the E-vector is in the plane perpendicular to the grating vector, the total diffraction efficiency is higher than if they are in the same plane. However, the larger total diffraction efficiency mainly results in higher intensity at the high diffraction orders, which means a little increase of the 0th order intensity.
For off-normal incidence (φ>0°), the total diffraction efficiency of s-polarized and p-polarized light increases and decreases, respectively, compared to its normal incidence case, due to the polarization effect of the ML reflection (Fig. 4). The 0th order efficiency of both s- and p-polarization slowly increases at longer wavelengths. This is mainly caused by the weakened blazed diffraction effect at larger φ as the angular dispersion of different orders is increased and the specular reflection angle from the facets is less matched to the diffraction angles. The changes of diffraction efficiency at off-normal incidence angles for 2D pyramids is similar as for the 1D case. Although the larger in-plane angle will increase the 0th order reflectance, it is still minor up to φ = 40°. The 0th order reflectance of un-polarized light at φ = 40°, θ = 0° is less than 10% over the entire range of λ = 100-400nm with an average value of 2.7% for 1D pyramids (h = 154nm, w = 0.04p). For the 2D pyramids (h = 105nm, w = 0.34p) mentioned above, the average reflectance of un-polarized light at φ = 40° is 7.7%.
2.4 EUV reflectance of the ML pyramids-on-mirror
The advantage of ML pyramids over Si pyramids is the higher EUV reflectance expected, contributing to a more efficient use of EUV photons generated. To illustrate such an effect, the EUV response around λ = 13.5nm of a 2D ML pyramids on top of a 50 bilayers ML mirror (MLM) system was calculated and the results are shown in Fig. 5.The pyramid structure is h = 105nm (containing 15 bilayers), p = a = 26µm, top width w = 0.34p, the incidence geometry is φ = 5°, θ = 0°, at p-polarization state. Fully coherent EUV light was assumed here. A 50 bilayer Mo/Si MLM was also calculated for comparison. As the EUV reflectance of a Mo/Si ML is saturated at 50 bilayers, the effect of extra bilayers above N = 50 can be neglected. No interface roughness or diffusion was taken into account in the calculations. It can be seen that the 0th order reflectance of ML pyramids is decreased due to the presence of other diffraction orders. However, the diffraction efficiency of higher orders drops very fast and the maximum efficiency of the 3rd orders is only 1.5%. If the diffraction of all orders from (−3,-3) to (3,3) (49 orders including the 0th) is summed up, it will give almost the same EUV reflectance profile as the unstructured MLM. The peak reflectance of the ML pyramids system is only 0.2% smaller than the MLM with a tiny reduction of the bandwidth. This small change of the reflectance profile is mainly caused by not including the other higher diffraction orders. According to the grating equation, , where α is the diffraction angle, the angular separation between the strongest 3rd orders ((3,0),(−3,0),(0,3),(0,-3)) and the 0th order at λ = 13.5nm is 0.09°. For the shortest wavelength of UV light, λ = 100nm, the separation between 0th and ± 1st order is 0.22°. This enables the collection of all the main diffraction orders of EUV light while still filter out the UV diffraction. For 1D pyramids, it has the same trend as 2D and the integrated diffraction efficiency is also only 0.2% less than the unstructured MLM. On the other hand, if the spatial coherence length of EUV is smaller than the period of pyramids, no diffraction orders will be generated and the EUV light will be simply reflected by the multilayer. It is worth to note that the polarization of EUV light or the E-vector orientation has little effect on the 0th order and integrated diffraction efficiency (at near normal incidence) for both 2D and 1D pyramids. This can be attributed to the fact that the EUV light can only be reflected by the periodic multilayer structure and the (non-)symmetric shape has little effect now.
3. First demonstration of the multilayer pyramids structure
Taking the dependence of UV suppression and EUV reflectance of the structural parameters into account, a proof of principle demonstration of 2D ML pyramids structure was designed and fabricated. The designed parameters are listed in Table 1.Please note that the designed ML pyramids is a very shallow tapered structure with a facet angle of only 0.7°. A flat Mo/Si ML with 50 bilayers and a d-spacing of d = 7.0nm was deposited as a reference using the e-beam evaporation technique with ion assistance . The fabricated ML pyramids structure was measured with atomic force microscopy (AFM) (Fig. 6) and its structural parameters are also listed in Table 1. The fabricated structure is consistent with the designed one while the height is corresponding to around 16 Mo/Si bilayers. The flat multilayer under the pyramids consists of more than 50 bilayers so that the EUV reflectance from the ML pyramids system is saturated.
4. UV and EUV measurements
Both the UV and EUV reflectance were measured at the Physikalisch-Technische Bundesanstalt (PTB) in Berlin, Germany. The reflectance of UV light was measured at 5° incidence angle (in plane) using p-polarized radiation. The beam divergence and the angular resolution of detector was reduced to separate the 0th reflection from the higher diffraction orders. As shown in Fig. 7, the flat Mo/Si ML mirror is highly reflective for the entire UV band with an average reflectance of 53%.For the ML pyramids structure, the entire UV band was fully suppressed and the average reflectance is only 3.7%. Thus, a 14 times average suppression was achieved. The UV reflectance at shorter and longer wavelengths can be further balanced by tuning the width of the top-flat and valley area, or increase the pyramid height. Moreover, a maximum suppression of over 400 times was obtained at λ = 135nm due to the destructive interference effect as shown in the inset graph. This maximum suppression can be shifted to a different wavelength by tuning the pyramid height which can be useful in case a specific certain strong emission line from the source needs to be suppressed. The theoretical reflectance calculated using the AFM measured structural parameters (Table 1) is close to the measured curve which illustrates the accuracy of the theoretical model. The small difference can be mainly caused by the deviation of the fabricated shape from the perfect 2D pyramid shape represented by the simplified structural parameters from AFM measurement.
The EUV reflectance was also measured at a 5° incidence angle with s-polarized light on the same sample. To collect all the diffracted power, the angular acceptance of the detector was first set at 1° ( ± 0.5°). The co-deposited flat Mo/Si MLM was measured for comparison. The peak reflectance of the flat ML mirror is 68.7%. For the ML pyramids on the MLM, a high reflectance of 64.7% is achieved, the highest value so far from different wavelength filtering devices, and only 4% absolute reflectivity loss. The relative EUV efficiency of this ML pyramids system is RML-pyra./RMLM = 94.2% which is much higher than has been reported e.g. for a transmission filter and the Si pyramids structure (also shown in Fig. 8) . Moreover, the peak wavelength position and bandwidth of the fabricated ML pyramids was almost the same as of the flat ML. (The minor difference of 10pm for the peak wavelength can be attributed to small differences between as-deposited samples). It is indicated that the d-spacing of the ML structure inside and under the pyramid was not affected during the fabrication. The small drop of reflectance of 4% can be caused by the damage and oxidation on the surface after the fabrication process.
To see the spatial distribution of the reflected EUV intensity, a CCD measurement was performed at PTB. The CCD has a pixel array of 2048 × 2048 with a pixel size of 13.5 × 13.5µm2. It was mounted at a distance of 1427mm from the sample. Due to the fixed position of CCD, the incident angle was set at 6.75° for the pyramid sample while the peak wavelength at this angle, λ = 13.5nm, was used for imaging. The measured CCD results are shown in Fig. 9 (the intensity is on a logarithmic scale). Given to the relatively good spatial coherence of the synchrotron radiation, a two dimensional diffraction pattern is observed which comes from the 2D ML pyramid structure. Nevertheless, the diffracted EUV intensity almost drops to the background noise level at ± 0.16° away from the peak intensity position, even along the strongest diffraction directions (x and y indicated in the image). The integral intensity of the center ± 0.16° angular range (indicated by the square box) amounts to 99.5% of the total intensity received by the CCD. As the angular acceptance of the whole CCD is a little larger than the photodiode used in the wavelength scan mentioned above, the integral intensity of the CCD corresponds to the 64.7% reflectance. This means that the unprecedented high EUV reflectance is within an angular range of ± 0.16°. The measured angular distribution of the reflected EUV intensity is a little wider than the ± 0.09° calculated for the ( ± 3, ± 3) orders range which can be attributed to the finite size and divergence of the incident beam and scattering from the fabricated ML pyramid structure. An angular scan with the detector in the UV reflectance measurement shows that the ± 1st order UV diffraction of λ = 120nm is located at ± 0.22° away from the 0th order and the longer wavelength will be diffracted to even larger angle. The angular distribution of both EUV and UV light can be further tuned by changing the lateral period of the pyramids structure. Thus, with the help of a slit, the fabricated pyramid-MLM structure can well separate the full band UV radiation from the EUV light with high efficiency.
A novel ML pyramid structure has been designed and fabricated on top of an EUV multilayer mirror for spectral purity enhancement. The demonstration optics has shown a full band suppression of the UV spectrum (λ = 100-400nm) with an average factor of 14 and a maximum suppression factor of over 400. As this spectral filter is solely made of a multilayer structure, it has in principle no loss of EUV reflectance and the demonstration optics shows an unprecedented high EUV reflectance of 64.7% near the target wavelength of 13.5nm. This corresponds to 94.2% of the reflectance of an unstructered, regular multilayer mirror, or an absolute loss of 4% only.
In addition to the 2D ML pyramids scheme used in this work, a comprehensive design showed that a 1D ML pyramids design can also achieve a similar suppression and efficiency. Both 2D and 1D ML pyramids can be applied to an incident angle of up to 40°. The 1D structure may have more advantages like easier implementation on large mirrors, depending on the fabrication method. Furthermore, this method can also be used for different wavelength ranges as long as the multilayer materials used are reflective for the unwanted longer wavelength. Thus, it can be applied to many other applications with broadband XUV sources like in solar imaging systems or optics for filtering high harmonics sources.
The authors are grateful to Dr. Christian Laubis and Dr. Alexander Gottwald at the Physikalisch Technische Bundesanstalt (PTB) in Berlin for the EUV and UV measurements. This work is part of the research program “Controlling photon and plasma induced processes at EUV optical surfaces (CP3E)” of the Stichting voor Fundamenteel Onderzoek der Materie. The CP3E program is co-financed by Carl Zeiss SMT GmbH (Oberkochen), ASML (Veldhoven), and the AgentschapNL through the Catrene EXEPT program.
References and links
1. I. V. Fomenkov, D. C. Brandt, N. R. Farrar, B. L. Fontaine, N. R. Böwering, D. J. Brown, A. I. Ershov, and D. W. Myers, “Laser produced plasma EUV light source for EUVL patterning at 20nm node and beyond,” Proc. SPIE 8679, 86792I (2013). [CrossRef]
2. Y. Teramoto, G. Niimi, D. Yamatani, Y. Joshima, K. Bessho, T. Shirai, T. Takemura, T. Yokota, H. Yabuta, K. C. Paul, K. Kabuki, K. Miyauchi, M. Ikeuchi, K. Hotta, M. Yoshioka, and H. Sato, “Development of Xe-and Sn-fueled high-power Z-pinch EUV source aiming at HVM,” Proc. SPIE 6151, 615147 (2006). [CrossRef]
3. F. Brandi, D. Neshev, and W. Ubachs, “High-order harmonic generation yielding tunable extreme-ultraviolet radiation of high spectral purity,” Phys. Rev. Lett. 91(16), 163901 (2003). [CrossRef] [PubMed]
4. J. Lilensten, T. Dudok de Wit, M. Kretzschmar, P.-O. Amblard, S. Moussaoui, J. Aboudarham, and F. Auchère, “Review on the solar spectral variability in the EUV for space weather purposes,” Ann. Geophys. 26(2), 269–279 (2008). [CrossRef]
5. E. Louis, A. E. Yakshin, T. Tsarfati, and F. Bijkerk, “Nanometer interface and materials control for multilayer EUV-optical applications,” Prog. Surf. Sci. 86(11-12), 255–294 (2011). [CrossRef]
6. I. A. Artyukov, A. I. Fedorenko, V. V. Kondratenko, S. A. Yulin, and A. V. Vinogradov, “Soft X-ray submicron imaging experiments with nanosecond exposure,” Opt. Commun. 102(5-6), 401–406 (1993). [CrossRef]
7. J. Fujimoto, T. Hori, T. Yanagida, T. Ohta, Y. Kawasuji, Y. Shiraishi, T. Abe, T. Kodama, H. Nakarai, T. Yamazaki, and H. Mizoguchi, “Development of laser-produced plasma-based EUV light source technology for HVM EUV lithography,” Proc. SPIE 8332(83220F), 83220F (2012). [CrossRef]
8. V. Y. Banine, K. N. Koshelev, and G. H. P. M. Swinkels, “Physical processes in EUV sources for microlithography,” J. Phys. D Appl. Phys. 44(25), 253001 (2011). [CrossRef]
9. N. I. Chkhalo, M. N. Drozdov, E. B. Kluenkov, A. Y. Lopatin, V. I. Luchin, N. N. Salashchenko, N. N. Tsybin, L. A. Sjmaenok, V. E. Banine, and A. M. Yakunin, “Free-standing spectral purity filters for extreme ultraviolet lithography,” J. Micro/Nanolith. MEMS MOEMS 11(2), 021115 (2012). [CrossRef]
10. M. M. J. W. van Herpen, R. W. E. van de Kruijs, D. J. W. Klunder, E. Louis, A. E. Yakshin, S. A. van der Westen, F. Bijkerk, and V. Banine, “Spectral-purity-enhancing layer for multilayer mirrors,” Opt. Lett. 33(6), 560–562 (2008). [CrossRef] [PubMed]
11. S. P. Huber, R. W. E. van de Kruijs, A. E. Yakshin, E. Zoethout, K.-J. Boller, and F. Bijkerk, “Subwavelength single layer absorption resonance antireflection coatings,” Opt. Express 22(1), 490–497 (2014). [CrossRef] [PubMed]
12. A. J. R. van den Boogaard, F. A. van Goor, E. Louis, and F. Bijkerk, “Wavelength separation from extreme ultraviolet mirrors using phaseshift reflection,” Opt. Lett. 37(2), 160–162 (2012). [CrossRef] [PubMed]
13. H. Kierey, K. Heidemann, B. Kleemann, R. Winters, W. Egle, W. Singer, F. Melzer, R. Wevers, and M. Antoni, “EUV spectral purity filter: optical and mechanical design, gratings fabrication, and testing,” Proc. SPIE 5193, 70–78 (2004). [CrossRef]
14. D. L. Voronov, E. M. Gullikson, F. Salmassi, T. Warwick, and H. A. Padmore, “Enhancement of diffraction efficiency via higher-order operation of a multilayer blazed grating,” Opt. Lett. 39(11), 3157–3160 (2014). [CrossRef] [PubMed]
15. Q. Huang, D. M. Paardekooper, E. Zoethout, V. V. Medvedev, R. W. E. van de Kruijs, J. Bosgra, E. Louis, and F. Bijkerk, “UV spectral filtering by surface structured multilayer mirrors,” Opt. Lett. 39(5), 1185–1188 (2014). [CrossRef] [PubMed]
16. Website of the Rosft, http://optics.synopsys.com/rsoft/rsoft-passive-device-diffractMOD.html
17. M. G. Moharam and T. K. Gaylord, “Rigorous coupled-wave analysis of grating diffraction - E-mode polarization and losses,” J. Opt. Soc. Am. 73(4), 451 (1983). [CrossRef]
18. D. Attwood, Soft X-rays and Extreme Ultraviolet Radiation, Principles and Applications (Cambridge University, 1999), pp 304–306.