We report on the development of true free-standing phase transmission gratings for the extreme ultraviolet band. An ultra-nanocrystalline, 300 nm thin diamond film on a backside etched silicon wafer is structured by electron-beam lithography to periods of 1 μm. In this way, flat and stable gratings of 400 μm in diameter are fabricated. First-order net efficiencies up to 28% are obtained from measurements at a synchrotron beamline within a wavelength range from 5.0 nm to 8.3 nm, whereas the 0th order is suppressed to 1% near 6.8 nm. Higher diffraction orders up to the 3rd one contribute less than 7% in sum to the far-field pattern.
© 2011 OSA
EUV and soft X-ray gratings represent important tools not only for high-resolution plasma spectroscopy in the laboratory [1, 2] and on space-borne astronomical instruments [1, 3], but also for the development of precise and efficient masks in the emerging field of short-wavelength interference lithography . In the past, blazed reflective optics have been built and tested as monochromators at a synchrotron beamline . Despite well-designed focusing properties by means of continuous variations of the grating period, the modest diffraction efficiencies obtained probably prohibit their implementation in low-signal applications, i.e. astronomy. Current techniques make still use of blazed devices for reflection  and “classical” amplitude transmission gratings , which may suffer from low misalignment tolerances , surface imperfections  or naturally limited diffraction efficiencies, regardless their precise fabrication . Recent developments also turned to transmissive phase-shifting molybdenum (Mo) versions [11, 12]. However, their performance is severely limited by solid substrates made of silicon nitride (Si3N4). In particular, an additional absorption of ≈ 58% caused by this membrane has to be accepted.
In this paper, we present true free-standing binary EUV phase transmission gratings for the spectral band beyond the carbon K-edge, fabricated from ultra-nanocrystalline (UNC) diamond films. We review general basics of diffractive EUV transmission optics in Sect. 2 and develop in Sect. 3 the specific design and thereon the technological manufacturing procedure as well. In Sect. 4, the experimental scheme for the measurements is described and the results are analyzed in detail. After a brief discussion of prospective applications in Sect. 5, we conclude with an outlook to further steps in Sect. 6.
2. Principles of phase transmission gratings with absorption
We consider a structure made of N ≫ 1 rectangular bars (B) and slits (S), whose periodicity is denoted by d. Its duty cycle or fill factor (FF) is defined as f ≡ b/d, where the quantity b < d represents the width of the bars. A schematic cross-section is shown in Fig. 1. The material properties in the EUV and X-ray range are well described by the complex refractive index n = 1 – δ – iβ, which accounts for both the weak refractive power – the typical case, near absorption edges or towards long wavelengths, deviations from this rule arise – by 0 < δ ≪ 1 and significant absorption. Within this work however, we use an equivalent two-dimensional set of parameters to characterize the spectral behavior of a given optical design. In particular,13], the “critical zone number” N 0 quantifies the number of π phase shifts within one absorption length. Hλ accounts for the normalized phase shift within the bars Bn via the π -thickness Δt π = λc/2δc. The symbol δc abbreviates the refractive index increment at a certain “design wavelength” λc. Using the definitions from Eq. (1) and based on scalar diffraction , the wavelength-dependent efficiency in an order m ≠ 0 is found as Eq. (2), which applies to all “thin” transmission gratings in the scalar approximation, Pm(λ) smoothly peaks for a fill factor f = 0.5 around λc, regardless of m ≠ 0. Moreover, an ambiguity with respect to complementary values, i.e. f and 1 – f, is incorporated in the factor Cm. Whereas higher orders with |m| ≥ 2 are strongly suppressed at least ∝ (πm)−2, phase gratings may concentrate the dominant fraction of the flux into the (±1)st orders for N 0 ≳ 1. In case of maximal and negligible absorption, i.e. 0 < N 0 < ∞, the (±1)st order efficiency is found as and , respectively. In the EUV, the best materials provide typical efficiencies P ±1 up to ≈ 32%. In contrast, the 0th order contribution to the diffraction pattern turns out to be quite sensitive to variations of the fill factor. We find Fig. 2 illustrates the (±1)st and the 0th order for realistic values of Hλ and N 0 in the EUV and soft X-ray band, specified for the fill factor of the sample in use (Sect. 3). As it is shown on the left of Fig. 2, the 0th order minimum with P 0(λ) ≤ 0.1% is strongly confined to Hλ ≈ 1.0 and 1.6 ≲ N 0 ≲ 2.3, an accessible region for λ ≲ 8 nm using high-purity diamond samples (see also Fig. 8 in Sect. 4). If matched for Hλ = 1 at λc, the power ratio between the (±1)st and the 0th order is found as Eq. (4), an infinite contrast would be reached for an exactly adapted fill factor f opt from above – however, inevitable fabrication errors will set limits to this singularity.
3. Design constraints and fabrication
Just beyond the K-edge near 4.36 nm, pure carbon excels in terms of its transparency with an absorption length up to 1.0 μm . Due to its density ≤ 3.52 g cm−3, diamond permits the lowest π-thickness Δtπ among all carbon phases, an advantage for an accurate application of the scalar theory – the question of its validity is addressed in the appendix – from Sect. 2 within that “proof-of-concept” work. In Fig. 3, the predicted performance based on Eq. (2) and Eq. (3) is shown in dependence on the fill factor and the thickness Δt, using the data from . Since the upper extremal regions with Δt > 0.6 μm refer to multiples of the π-thickness, we select the lowest one to obtain the highest (±1)st order efficiency. In this way, the recommended design is given by f opt = 0.556 and λc = 5.89 nm for Δt = 0.31 μm. This thickness is indeed the fixed parameter of the diamond film in use: We use polished, ∼ 500 μm thick 4” Si wafers in 〈100〉 orientation, coated with a (0.3 ± 0.06) μm thin UNCD film whose density is specified to 3.30 g cm−3 . By means of scanning wide-band reflectometry in the ultraviolet and visible regime for a 6% decrease  with respect to the nominal refractive index in this wavelength band , the thickness Δt of the unstructured Diamond-on-Silicon (DoSi) layer at the position of the future grating sample is initially determined with an accuracy of ±2%.
The grating fabrication is divided into front- and backside-structuring. As shown in Fig. 4, we start with the etching process of the diamond layer. A metal hard mask, 50 nm chromium in this case, is coated by magnetron sputtering. On top, 300 nm of the chemically amplified electron beam resist FEP 171 is spin coated. Therefore, a low electron dose of 9.5 μC cm−2 is sufficient. With a variable shaped beam electron beam writer (Vistec SB 350 OS), a large area with sub-micron structures can be exposed rapidly. The dry etching of the chromium mask is done with reactive ion etching (RIE) in a conventional parallel-plate reactor SI-591 (Sentech Instruments, Berlin) by means of a chlorine-oxygen plasma. The diamond etching itself is done in the same plasma reactor (SI-591) and stops perfectly on the Si substrate.
At this point, the front side structuring is finished. The chromium on top of the gratings remains tentatively for protecting the diamond gratings in the following steps of backside structuring. Now the Silicon is etched from the backside through the wafer to the diamond layer on the front side towards self-supporting diamond membranes. To form circular, mm-sized windows in the Si wafer, approximately 14 μm of the DNQ-Novolak based standard photoresist AZ 4562 (Clariant) is spin coated on the backside of the wafer. After soft baking on a hot plate, the resist is exposed with a photo mask in an EVG mask aligner AL 6-2. Due to the large structures in the mm range, there is no need for high resolution. The 14 μm AZ resist layer is fully sufficient to etch more than 700 μm in silicon. The residual chromium on top of the diamond gratings is now removed by RIE. Afterwards, the trough-wafer etching (VIAS) is done by ICP-RIE etching based on the BOSCH® process in a SI-500 C etching tool (Sentech Instruments) at 300 K. Since the Si etching process stops quite well on the diamond, no separate etch stop layer is required for this final step.
A series of free-standing, about 1.4 mm wide membranes and inscribed gratings with a period of 1.0 μm and fill factors 0.55 ≲ f ≲ 0.63 is obtained. To avoid stress-induced irregularities in the grating topography and an associated optical derogation, we choose a circular shape with perpendicular stabilizers (50 μm period), located in the central low stress area of the membrane. SEM pictures are shown in Fig. 5. From the incorporated fill factor variation with an increment of ≈ 10 nm, the best match to the value for f opt from above is found as f emp = 0.564 ± 0.006; this sample is selected for the measurements described in Sect. 4.
4. Measurement technique and results
In normal incidence transmission, all measurements are based on the angular-dependent far-field diffraction pattern recorded in a distance of 980 mm from the sample by a photo diode whose entrance slit width is specified to 1 mm. Filtered by a monochromator to a relative bandwidth of ≲ 10−3 and collimated to few 102 μm, efficiencies are determined from the normalized transmission of the TE-polarized synchrotron beam with a nearly symmetric two-dimensional Gaussian intensity profile g(r⃗). In particular, its full widths at half maximum (FWHM) are found from edge scans as (206 ± 2)μm and (196 ± 5)μm. Centered to the grating sample with an accuracy of ±10μm using the local 0th order transmission minimum, the weighted coverage of the unstructured diamond membrane by the Gaussian beam amounts to 15.7%, including both the stabilizing grid and the surrounding area. In order to deduce the contribution of this non-diffracted light to the 0th order, the wavelength-dependent transmission through an unstructured stand-alone membrane is measured elsewhere on the DoSi wafer. From its thickness ratio to the sample film – evaluated via reflectometry again – valid results can be extrapolated to T 0(λ) for the grating sample in use, as listed in Tab. 1. An estimation of the artificial 0th order contribution P̃ 0(λ) from non-grating regions covered by the beam is straightforward now,Eq. (2) to receive the net grating performance. Table 2 gives an overview on both data sets for the diffracted efficiency P ±m with 1 ≤ |m| ≤ 3 within the scanned wavelength region from 5 nm to 7.5 nm. As justified by the measurements, almost equal results from the negative and positive orders are averaged in each case, . The (±1)st order clearly dominates with corrected efficiencies beyond 20%. In contrast, the (±2)nd order drops to almost negligible values < 1%, whereas the (±3)rd order yields ∼ 2%, in good agreement with Eq. (2). Due to its particular importance, the 0th order is measured within an extended band from 5 nm to 8.3 nm. The results are presented in Fig. 6. We detect an almost perfect collapse of the corrected data at 6.77 nm to P 0 = 1.0%, an indicator for the good optical quality of our gratings. The error bars shown in Fig. 6 refer to the uncertainty in the artificial 0th order contribution P̃ 0(λ) which originates from the standard deviation of the FWHM beam diameters from above. On the right of Fig. 6, the power ratio between the (±1)st and the 0th order is shown, with the same error estimation as before. Since this contrast is mainly driven by the 0th order, the peak is found for the same wavelength, which may be identified with the empirical result for λc. Near this central wavelength, the raw data for reach their – original – minimum at 6.85 nm. Fig. 7 shows the far field pattern recorded at this wavelength for both the raw and corrected data set again. The efficiencies detected in this scan agree quite well with the results from Tab. 2. However, the 0th order intensity differs by 0.4% – an aberration which might be attributed to slight alignment tolerances between those independent measurements. Compared to Fig. 3, λc is shifted towards longer wavelengths by about 1 nm. This effect may be explained by deviations of the intrinsic optical properties of the diamond sample, namely δ(λ), from its tabulated values. We apply semi-analytical best-fit procedures which adjust Eq. (2) and Eq. (3) to the corrected data using appropriate parameters for Hλ and N 0. In conjunction with the measured thickness Δt from above, the complex refractive index parameters of the diamond sample can be derived, as shown on the left in Fig. 8. In this case, the statistical uncertainty is negligible and error bars have been omitted, although systematic discrepancies might arise, based on the non-ideal shape of P 0(λ) around its minimum. From its definition, the critical zone number N 0 is now directly obtained from those estimations for δ and β, as illustrated on the right of Fig. 8. The error bars in this plot rely on the proximate 2% tolerance in the measured thickness Δt.
That moderate drop-down of N 0 by ∼ 15% is typical for “real” high-purity materials whose underlying refractive index data are close to the ideal ones. An investigation of its origin would be nonetheless worthwhile. The manufacturer of the DoSi sample claims negligible metallic contaminations like 13Al, 26Fe and 74W on the ppb level or below . Potential sources of refraction and transmission losses might be rather found in low-Z elements like nitrogen (7N) and oxygen (8O), due to their nearby absorption edges near 3.0 nm and 2.3 nm, respectively.
5. Towards applications
The results from above should be regarded as a proof-of-principle in the subject of stand-alone EUV phase transmission gratings. Compared to reflective optics, such components are usually favored with respect to their weight and alignment tolerances – advantages which are of particular interest in astrophysics. Significant progress in plasma spectroscopy requires a resolving power λ/Δλ of at least 104 , associated with grating periods in the order of ∼ 100 nm for typical beam diameters around 10−3 m and aspect ratios up to ∼ 10. If properly designed, a good (±1)st order performance would be still maintained for “thick” structures, as it is shown in Tab. 3. Serious but not insuperable technical challenges come along with decreasing grating periods; amongst others, an improved etching procedure has to be developed. The 0th order is usually not in the scope of interest for spectroscopy. However, 0th and higher (|m| ≥ 2) order contributions degrade the contrast in interference lithography . Their suppression is thus considered as one of the main tasks in the design of useable EUV phase masks. For the sample from Tab. 3, the 0th order decreases to ≤ 0.01% near 6.6 nm whereas all orders with |m| ≥ 2 sum up to 7.4%. Advanced gratings are currently investigated in our research group which are capable to diminish this amount of light and provide a resolution down to few 10 nm by frequency doubling . For instance, variations of the binary profile with a period of ∼ 60 nm reduce the summed contribution for |m| ≠ 1 to 2% – with a (±1)st order efficiency up to 34%.
In summary, we fabricate and test free-standing binary phase transmission gratings made of commercially available diamond films for the EUV band with a (±1)st order efficiency up to ≈ 28%. An absorbing support layer is avoided, making optics of that type suitable for high-efficient spectroscopy on board of future space-borne missions or for plasma diagnostics in the laboratory. The suppression of the 0th order to 1% further suggests an application to near-field interference lithography. Next research steps should optimize the optical performance and reduce the grating period to 0.1 μm or less to receive an ultra-high spectral and spatial resolution. Moreover, their astronomical usage requires large-scale segmented assemblies for which robust low-stress ultra-thin membranes should be combined with an appropriate support grid.
Within this work, we use the scalar theory which permits an elegant access to the analysis of absorbing phase gratings. Deviations to an exact RCWA calculation arise for “thick” gratings and small feature sizes, compared to the wavelength . In the following, we investigate how the period affects the efficiencies Pm(λ) for the design from Sect. 3. Since the error between the RCWA model ( ) and the scalar approach ( ) strongly varies with λ we define a universal quantity σPm to characterize the wavelength-independent discrepancy viaFig. 9. As expected, the mean error significantly increases towards small periods. For d = 1.0 μm, we obtain σPm ≲1% for all orders, an acceptable level for sufficiently accurate results. Table 4 gives an explicit overview for this chosen feature size. Aside from the relatively small efficiency errors, the positions of the 0th order minimum – a sensitive mismatch indicator –differ by 0.036 nm, which is within the uncertainty of 0.050 nm for the acquired data sample.
This work was part of the project “Zentrum für Innovationskompetenz ultra optics”, funded by the German Federal Ministry of Education and Research (BMBF) with the fund number 03Z1HN32. All EUV measurements have been performed by F. Scholze, C. Laubis, C. Buchholz and coworkers from the PTB at the synchrotron radiation facility BESSY II in Berlin.
References and links
1. M. L. Schattenburg, E. H. Anderson, and H. I. Smith, “X-ray/VUV Transmission Gratings for Astrophysical and Laboratory Applications,” Phys. Scripta 41, 13–20 (1990). [CrossRef]
2. C. Peth, F. Barkusky, and K. Mann, “Near-edge x-ray absorption fine structure measurements using a laboratory-scale XUV source,” J. Phys. D: Appl. Phys. 41, 105202 (2008). [CrossRef]
3. R. K. Heilmann, M. Ahn, A. Bruccoleri, C.-H. Chang, E. M. Gullikson, P. Mukherjee, and M. L. Schattenburg, “Diffraction efficiency of 200-nm-period critical-angle transmission gratings in the soft x-ray and extreme ultraviolet wavelength bands,” Appl. Opt. 50, 1364–1373 (2011). [CrossRef] [PubMed]
4. H. H. Solak, “Nanolithography with coherent extreme ultraviolet light,” J. Phys. D 39, R171–R188 (2006). [CrossRef]
6. H. Lin, L. Zhang, L. Li, C. Jin, H. Zhou, and T. Huo, “High-efficiency multilayer-coated ion-beam-etched blazed grating in the extreme-ultraviolet wavelength region,” Appl. Opt. 49, 4450–4459 (2010).
7. D. R. McMullin, D. L. Judge, C. Tarrio, R. E. Vest, and F. Hanser, “Extreme-ultraviolet efficiency measurements of freestanding transmission gratings,” Appl. Opt. 43, 3797–3801 (2004). [CrossRef] [PubMed]
8. K. Flanagan, M. Ahn, J. Davis, R. Heilmann, D. Huenemoerder, A. Levine, H. Marshall, G. Prigozhin, A. Rasmussen, G. Ricker, M. Schattenburg, N. Schulz, and Y. Zhao, “Spectrometer concept and design for X-ray astronomy using a blazed transmission grating,” in Optics for EUV, X-Ray, and Gamma-Ray Astronomy III, Stephen L. O’Dell and Giovanni Pareschi, eds., Proc. SPIE6688, 66880Y (2007).
9. F. Salmassi, P. P. Naulleau, E. M. Gullikson, D. L. Olynick, and J. A. Liddle, “EUV Binary Phase Gratings: Fabrication and Application to Diffractive Optics,” http://escholarship.org/uc/item/6d42k5t7 (2005).
10. P. Predehl, H. W. Braeuninger, A. C. Brinkman, D. Dewey, J. J. Drake, K. A. Flanagan, T. Gunsing, G. D. Hartner, J. Z. Juda, M. Juda, J. S. Kaastra, H. L. Marshall, and D. A. Swartz, “X-ray calibration of the AXAF Low Energy Transmission Grating Spectrometer: effective area,” in Grazing Incidence and Multilayer X-Ray Optical Systems, R. B. Hoover and A. B. C. Walker II, eds., Proc. SPIE3113, 172–180 (1997).
11. P. P. Naulleau, C. H. Cho, E. M. Gullikson, and J. Bokor, “Transmission phase gratings for EUV interferometry,” J. Synchrotron Rad. 7, 405–410 (2000). [CrossRef]
12. M. Saidani and H. H. Solak, “High diffraction-efficiency molybdenum gratings for EUV lithography,” Microel. Eng. 86, 483–485 (2009). [CrossRef]
13. B. X. Yang, “Fresnel and refractive lenses for X-rays,” Nucl. instr. in Phys. Res. A 328, 578–587 (1993). [CrossRef]
14. C. Braig, P. Predehl, and E.-B. Kley, “Efficient extreme ultraviolet transmission gratings for plasma diagnostics,” Opt. Eng. 50(6), 066501 (2011). [CrossRef]
15. B. L. Henke, E. M. Gullikson, and J. C. Davis, “X-ray interactions: photoabsorption, scattering, transmission, and reflection at E=50-30000 eV, Z=1-92,” Atomic Data and Nuclear Data Tables 54(2), 181–342 (1993). [CrossRef]
16. Diane P. Hickey, Advanced Diamond Technologies Inc., 429 B Weber Road, #286, Romeoville, IL 60446, United States, http://www.thinDiamond.com (personal communication, 2010).
17. M. Bass, C. DeCusatis, G. Li, V. N. Mahajan, and E. van Stryland, Handbook of Optics: Optical properties of materials, nonlinear optics, quantum optics (McGraw Hill Professional, New York, 2009).
18. K. S. Wood, M. P. Kowalski, R. G. Cruddace, and M. A. Barstow, “EUV spectroscopy in astrophysics: The role of compact objects,” Adv. Space Res. 38, 1501–1508 (2006). [CrossRef]
19. D. A. Pommet, M. G. Moharam, and E. B. Grann, “Limits of scalar diffraction theory for diffractive phase elements,” J. Opt. Soc. Am. A 11, 1827–1834 (1994). [CrossRef]