Open Access
Add to BibSonomyAbstract
We report on a dual-wavelength diffractive beam splitter designed for use in parallel laser processing. This novel optical element generates two beam arrays of different wavelengths and allows their overlap at the process points on a workpiece. To design the deep surface-relief profile of a splitter using a simulated annealing algorithm, we introduce a heuristic but practical scheme to determine the maximum depth and the number of quantization levels. The designed corrugations were fabricated in a photoresist by maskless grayscale exposure using a high-resolution spatial light modulator. We characterized the photoresist splitter, thereby validating the proposed beam-splitting concept.
© 2016 Optical Society of America
1. Introduction
Diffractive beam splitters (DBSs) are optical beam-delivery elements based on light diffraction. They are widely used in laser-material processing for the purpose of increasing productivity in drilling, cutting, and joining [1–5]. By using a DBS with a focusing lens, a laser beam can be efficiently divided into an arbitrary array of beam spots to process a workpiece at multiple points, thereby increasing the throughput. Meanwhile, using two different wavelengths simultaneously to irradiate a material can increase the scope and quality of processing [6–9]. In traditional laser-processing applications, a specific wavelength is usually chosen. The prospect of improving the performance of dual-wavelength processing motivated us to develop a dual-wavelength DBS for parallel laser processing. This novel splitter, made from a transparent material such as resist, glass, or polymer, generates two beam arrays with different wavelengths and allows them to coincide at the process points on the workpiece.
DBSs operating as designed at several wavelengths have been studied in the context of laser displays, spectroscopy, data communications, etc [10–15]. Such splitters with quantized surface-relief profiles were designed using iterative algorithms; Gerchberg-Saxton [10, 11], direct binary search [13, 15], and optimal-rotation angle [12, 14]. The cost functions in the design were extended to deal with multiple wavelengths. Regarding the design of a dual-wavelength DBS, it is generally accepted that the surface phase profile must be deep enough to cater for the two spectral components. The underlying idea, as explained by Noach et al. [10], involves continually adding integer multiples of 2π to a phase at one wavelength until the phase at the other wavelength acquires a desired value. Using this idea, Barton et al. [11] experimentally demonstrated the generation of two-dimensional light patterns using a two-color DBS with a 16-level relief profile, operating with blue and red light. The idea is easy to implement; yet, it imposes constraints on the choice of phases and also causes phase errors. Phase fitting is not necessarily guaranteed because of fabrication-related limitations to the depth range and resolution. There has therefore been interest in developing an alternative scheme for finding an appropriate combination of maximum depth and number of quantization levels.
In this study, we designed a dual-wavelength DBS for combining green and the near-IR, two versatile wavelength regimes in laser materials processing. We used a simulated-annealing (SA) algorithm to optimize the deep phase profiles with multiple levels. Owing to their capability of escaping from local minima, SA algorithms in general have a good chance to reach one of the best solutions close to the global minimum. We introduced a simple and practical scheme to select the maximum phase and the number of phase levels, thereby increasing the likelihood of finding a solution by iterative design. With the obtained profile, the error sensitivity was analyzed by incorporating profile errors to determine the fabrication strategy. The designed splitter was fabricated from photoresist, using a maskless grayscale exposure process based on a high-pixel density digital micromirror device to avoid sacrificing resolution through the repeated use of photomasks. We finally evaluated the optical properties of the splitter, to validate the proposed beam-splitting concept.
2. Concept
Figure 1 outlines the concept of the laser-processing system with the dual-wavelength DBS. (The diffraction angles are enlarged for clarity.) Two laser beams of different wavelengths are combined with a beam combiner before entering the splitter. Each beam is split into equally intense beams by the splitter, and the split beams are delivered through an achromatic lens onto a beam array. Combined with a XY stage or a galvano scanner, a one-dimensional splitter can cover large working areas. The splitter performance is characterized by the beam-splitting efficiency and uniformity. The efficiency measures the proportion of the input power that becomes focused into the beam array, while the uniformity equals the ratio between the minimum and maximum intensities among the split beams. This system allows the two beam arrays to overlap and irradiate the process points on the workpiece. Beam registration requires that the two wavelengths satisfy a specific relation, as described below. The split-beam locations on the focal plane are given by x1 = m1λ1f/p and x2 = m2λ2f/p, where the subscripts denote the wavelength, m1 and m2 are the diffraction orders, f is the lens focal length, and p is the splitter period. As f is identical for λ1 and λ2, we have
The combination of the two laser wavelengths allows variety in terms of the laser-emission mode (CW or pulsed) and light-matter interaction, and provides a range of output powers. For example, if λ1 and λ2 are chosen to be a fundamental wave and a second harmonic wave, respectively, then Eq. (1) becomes 2m1 = m2. Some undesired diffraction orders may therefore occur alongside the desired ones and should contain no energy to avoid damaging the workpiece. This is an important constraint incorporated in the design of dual-wavelength DBSs.We designed a Fourier-type splitter to produce the beam arrays in the far field. For near-field illumination, a Fresnel-type splitter can be designed by adding a lens-like function to the beam-splitting phase function. However, a Fresnel-type splitter lacks flexibility in terms of its ability to focus the beam arrays of different wavelengths onto a single plane, which severely limits their use in laser-processing applications. Instead of using two stand-alone lasers with different operating wavelengths, a dual-emission laser that emits two mutually coherent beams with different wavelengths along the same axis can be used in this system. Spatial interference between the beams is not problematic because they can be polarized in orthogonal directions by using a waveplate. That polarization control requires the splitter performance to be independent of the incident polarization, which is attainable.
3. Design
The dual-wavelength DBS design is based on a multiple-level transmissive, surface-relief periodic grating. As shown in the inset in Fig. 1, one period was unequally divided into zones, each with a given phase, to define a quantized surface-relief profile. A high-performance dual-wavelength DBS needs more zones and levels than a conventional splitter designed for a specified wavelength. Because of competition between the diffraction properties for the two wavelengths, the profile must be tailored to achieve the best compromise.
The surface profiles were stochastically designed using a SA algorithm [16] by considering the wavelengths, material dispersion, and splitting performance. A randomly chosen initial profile was iteratively optimized by perturbing the zone widths and phases. The cost function F was defined as
where m is the diffraction order, Im is the m-th order diffraction intensity, α1m and α2m are the target intensities, and ω1 and ω2 are weighting parameters. The summation was taken over the orders of interest, including the signals and noises. The weights were introduced to control the contributions to the splitter performance from the two wavelengths. Whether a solution can be found or not depends on these weights, as well as on the initial profile and the temperature scheduling. “Temperature” here means a parameter that regulates the probability of accepting negative perturbations. With decreasing annealing temperature, the probability also decreases, leading the design into a convergence. Using Fourier analysis [17], the diffraction intensities were computed aswhere N is the number of zones, ϕn (λ1) and ϕn (λ2) are the quantized phases at the n-th zone, and xn is the position of the phase transition. The splitter period was normalized to unity. Assuming that the period is much greater than the wavelengths and the splitter depth, the behavior of the splitter becomes insensitive to the polarization status of the incident light [18], and thus a scalar approximation applies to the computation of the diffraction intensities. This assumption holds for DBSs used in laser microprocessing, as they belong to Raman Nath regime with their Q values of ~0.001. If λ1 is chosen as the primary wavelength, ϕn (λ1) and ϕn (λ2) are chosen randomly from the following sets of phases, ϕ1k and ϕ2k, respectively,where ϕmax is the maximum phase depth, K is the number of phase levels, and n1 and n2 are the refractive indices of the splitter substrate at λ1 and λ2, respectively.A pair of weight values in Eq. (2) can balance the efficiencies and uniformities between the wavelengths. Figure 2 shows an example design of a 15-fan-out fused silica splitter which operates at 1064 and 532 nm. These wavelengths are relevant in laser processing, as they cause distinctive absorption in materials of interest. The efficiencies and uniformities are plotted against the annealing temperature, where the lines were smoothed for clarity. The design conditions were set as follows: N = 38, ϕmax = 2π, K = 30 with ω1:ω2 = 2:1 and α1m = α2m = 0.059 for the signals. These target intensities were set by considering a balance between the efficiency and uniformity. Using higher targets, the efficiency would be prioritized. The refractive indices were calculated using the Sellmeier series formula [19]; n1 = 1.450 at 1064 nm and n2 = 1.461 at 532 nm. As the temperature decreased, the efficiencies saturated above 80%, and the uniformities reached 0.85 or higher at both wavelengths. This design requires a stronger weighting for the fundamental wave than for the second harmonic. The window of weight ratios that yields a solution was also rather narrow because of the thin relief (ϕmax = 2π). This window was widened by deepening the relief profile, as explained below.
Fig. 2 Example design using a SA algorithm: (a) efficiencies and (b) uniformities. The split count is 15, the number of zones is 38, the maximum phase depth is 2π, the number of phase levels is 30, and ω1:ω2 = 2:1.
We introduced the following simple scheme to determine ϕmax and K. Starting from a given value of ϕmax (or K), K (or ϕmax) is varied until the values of mod (ϕ1k, 2π) and mod (ϕ2k, 2π) become evenly distributed between 0 and 2π. These values of ϕmax and K should be chosen by considering fabrication constraints of resolution and grayscale. This scheme provides several choices of ϕ2k for a given choice of ϕ1k, or vice versa. Taking the design in Fig. 2 as an example, Fig. 3 shows ϕ1k and ϕ2k plotted against k over the range 0 to 29 for different values of ϕmax. For ϕmax = 2π (Fig. 3(a)), only one value ϕ2k can be chosen for a given ϕ1k. For ϕmax = 6π (Fig. 3(b)), three values of ϕ2k can be chosen, and for ϕmax = 10π (Fig. 3(c)), five choices of ϕ2k are available. On the other hand, the phases are almost evenly spaced between 0 and 2π for ϕmax = 2π and 6π, but not for ϕmax = 10π. Using these values of ϕ1k, ϕ2k, and ϕmax, the coste functions were computed for 13 different pairs ω1:ω2 = 2.2:1, 2:1, 1.85:1, 1.66:1, 1.5:1, 1.22:1, 1:1, 1:1.22, 1:1.5, 1:1.66, 1:1.85, 1:2, and 1:2.2, where ω1 + ω2 = 2.0. The results are shown in Fig. 4 with the curves smoothed for clarity. The cost functions converge below 0.001 (a visual guide) for one pair of weights with ϕmax = 2π, nine pairs with ϕmax = 6π, and four pairs with ϕmax = 10π. With the function saturating at a lower value, we have a better chance of obtaining a profile that satisfies the performance requirements. From the results in Figs. 3 and 4, we conclude that the degree of freedom for design increases, only if a large phase depth is associated with mod (ϕ1k, 2π) and mod (ϕ2k, 2π), spaced uniformly over the range of 2π. The combination of ϕmax = 6π with K = 30 is just one of many appropriate combinations. The above-mentioned scheme is heuristic but practical for the iterative design of a dual-wavelength DBS. As regards the relation between K and N, we learn that K can be decreased by increasing N.
Fig. 3 Phases assigned to two wavelengths, 1064 nm (yellow) and 532 nm (green), as a function of the kth-level: the maximum phase depth is (a) 2π, (b) 6π, and (c) 10π.
Fig. 4 Cost function vs. annealing temperature with 13 different pairs of weights: The maximum phase depth is (a) 2π, (b) 6π, and (c) 10π.
The fabrication of a DBS requires converting the designed phases to real depths as
Note that the real depths are related to the longer wavelength. It is obvious from Eq. (5) that a large ϕmax yields a large minimum step because K can be chosen independently of ϕmax. As a proof of concept, we chose to form the splitter using a positive photoresist (AZ P4620, Electronic Materials), as opposed to glass or a polymer that require additional processes such as dry etching and imprinting, which may introduce profile errors. The refractive indices of the photoresist were calculated as functions of wavelength using the Cauchy series formula n = A + B/λ2 + C/λ4 [20], where A = 1.6207, B = 2.9136 × 10−3μm2, and C = 2.78 × 10−3μm4 are the Cauchy coefficients of the bleached resist (Sellmeier coefficients were not available) [21].We chose a small number of split counts to minimize the effects of depth errors on the splitter performance. Figure 5 shows a designed profile for five split counts for wavelengths 1064 and 532 nm. The design conditions were N = 40, ϕmax = 8π, and K = 22, with n1 = 1.625 at 1064 nm and n2 = 1.666 at 532 nm. The final optimized profile comprised 24 zones and 11 levels. The period was set to 1 mm, large enough to reduce the effects of fabrication errors and to verify the device concept. The chosen weights were ω1 = 1.2 and ω2 = 0.8. The minimum width and the minimum step in this profile were 10.0 μm and 1.54 μm, respectively, and Eq. (3) yielded the maximum depth as 6.19 μm. The theoretical efficiencies and uniformities were, respectively, 80% and 0.97 at 1064 nm, and 82% and 0.94 at 532 nm. The profile in Fig. 5 shows reflection symmetry, which halved the computation time in the design. The profile was designed in approximately 15 min using a desktop computer with a CPU clock rate of 3.3 GHz. The relief profiles can be optimized for different wavelength combinations, e.g., a fundamental wave and a third harmonic or a fourth harmonic, by utilizing the presented scheme to provide a set of appropriate phases for both wavelengths.
4. Error analysis
The effects of profile errors (round-off and depth errors) on beam-splitting performance were investigated with the 5-fan-out splitter profile in Fig. 5. Round-off errors were due to rounding off the phase transition positions to create pattern data for exposure, whereas depth errors were unavoidable in resist patterning. The results are shown in Figs. 6 and 7. In the simulations, the errors were added to the designed phase profile by rounding off xn and scaling ϕmax in Eqs. (3) and (4), respectively, and the diffraction intensities were computed using a discrete Fourier transform. The round-off error in Fig. 6 is defined as the ratio of the round-off size to the splitter period; the round-off size is determined by our resist-patterning technique. The scaled depth 1.00 in Fig. 7 corresponds to the designed depth. As seen from the simulations, (1) the profile errors tended to influence the uniformities more than the efficiencies and the shorter wavelengths more than the longer ones; (2) tolerances for the errors were approximately ~1% (corresponding to ~10 μm laterally and ~60 nm depth-wise), to ensure an efficiency > 80% and a uniformity > 0.80. We also studied random depth errors that would occur as a result of surface roughness and nonlinear resist characteristics. We found that they would be as detrimental as scaling errors. Errors in computing the refractive indices scale the depth up and down, thereby shifting the lines in Fig. 7 laterally.
Fig. 6 Beam splitting performances vs. round-off error (computer simulations). The plots are the simulation results and the lines are a guide for the eye.
To overcome tight tolerances, we can fabricate resist patterns on a substrate with intentionally varied exposure dose, and then select a sample that satisfies the performance requirements for use or profile-error criterions for replication. This approach, although costly, may be acceptable in cases where the splitters are intended for industrial applications such as laser processing. We can also compare several solutions and select a profile that is least error-sensitive. The relation between a profile and its error sensitivity needs to be investigated.
5. Fabrication
The fabrication of the designed splitter requires a high resolution (with a round-off error below 1%) and grayscale comprising more than 100 levels, as indicated by the error analysis. Accordingly, we employed maskless optical lithography using a digital micromirror device (DMD) [22]. Though DMS is a binary (ON/OFF) device, the durations of ON time and OFF time at each pixel can be controlled precisely and rapidly through input grayscale signals to achieve grayscale intensity modulation. The fabrication process is as follows: (1) the 400-cp photoresist was spin-coated at 1500 rpm for 60 s to a thickness of ~8 μm on a 1-mm-thick 100-mm-diameter fused silica substrate; (2) after pre-baking at 110°C for 10 min on a hotplate, a 10 mm × 10 mm area of the resist was exposed using a maskless exposure system (DL-1000/NC2P, NanoSystem Solutions) equipped with a UV-LED light source and a DMD by scanning the resist substrate in the XY directions; (3) after exposure at 405 nm, the resist was developed with 2.38%-TMAH at room temperature for 5 min and then rinsed in distilled water for 1 min to form the multi-level pattern. Before coating with resist, the substrate surface was treated with HDMS to promote resist adhesion. The resist thickness was measured to be 8.69 ± 0.13 μm across the substrate with a ~2.5-μm-thick resist layer left between the corrugations and the substrate. The patterns were not post-baked to avoid deforming the corrugations.
To characterize the resist, a 50-level test pattern was formed in it, and the profile was measured using a 3D laser microscope (OLS 4000, Olympus). The gamma characteristic curves obtained are shown in Fig. 8, where the circles are the measured depths before gamma correction, and the squares are the counterparts after gamma correction. A microscopic image of the test pattern is presented in the inset. The dose data for linearization was obtained by approximating the non-linear characteristic curve with a second-order polynomial equation. In this study, this dose correction was essential to decrease the depth errors, in particular, at low doses, corresponding to corrugations with a depth < ~2 μm. A simple and reliable linearization model is presented in Ref. 23.
The bitmap-formatted pattern data for the 5-fan-out splitter in Fig. 5 were displayed on the DMD with 1024 × 768 pixels, a pixel size of 14 μm × 14 μm, and 256 grayscale levels, and were projected onto the resist with a 1 μm × 1 μm pixel resolution, i.e., a 0.1% round-off error. Given this small round off, the influence of the DMD pixelation was negligible. Twenty five patterns, each of area 10 mm × 10 mm, were formed on a substrate by varying the exposure dose from 209 to 221 mJ/cm2 to complement the small tolerance of the scaling errors. The dose increment was set to 0.5 mJ, corresponding to ~15 nm in the etched depth, smaller than the depth-error tolerance. The exposure time was approximately 6 min on average for each pattern consisting of 13 × 10 shots, i.e., 2.8 s for each shot.
6. Evaluations
Only five patterns produced with the exposure dose range from 209 to 211 mJ/cm2, corresponding to a ~60 nm depth range, fanned out five beams at each of the two wavelengths. Because of its relatively high performance, the surface of the 210.5-mJ/cm2 pattern was profiled using the 3D laser microscope. A captured image is shown in Fig. 9, where the scales are different between the vertical and horizontal axes. The observed area is marked on the designed profile in Fig. 5. The profile depths were measured at 12 selected points with respect to an unexposed zone in half a period because of the profile symmetry (see the circled numbers in Fig. 9). The formed profile was deeper than the designed one at every point, indicating that the profile was somewhat scaled. The scaling factor was identified as being between 1.02 and 1.15, and tended to be larger at the shallow points (points 4, 9, and 12 in Fig. 9). This scaling variation is attributed to residual nonlinearity in the gamma characteristics, while some nonuniformity on the DMD is also suspected. To identify DMD-induced errors, we exposed the resist with a constant dose set across the entire projection area. By profiling this sample, the maximum depth (point 6 in Fig. 9) was found to be 6.39 μm, and the depth errors due to the DMD were estimated to be ± 70 nm (peak to peak) by taking account of the uneven resist thickness and the surface roughness. The surface roughness was found to be ± 60 nm (peak to peak) in the exposed area and ± 5 nm in the unexposed area. These errors degraded the beam-splitting performance, as described below.
Fig. 9 Measured surface profile of the splitter compared to the design: Circled numbers indicate the zones in which depths were measured.
The diffraction properties of the splitter were evaluated using the experimental setup outlined in Fig. 10. It contains two linearly polarized CW lasers, operating at 1064 and 532 nm with 20 mW outputs and TEM00 modes, and a beam profiler (BeamMic, Ophir-Spiricon) with a wide spectral response covering the range 350 to 1100 nm and a pixel size of 4 μm × 4 μm, to capture the two arrays of five split beams. The splitter was placed downstream of a focusing lens so that the array size could be varied by positioning it between the lens and the profiler. The lens was not achromatic and its focal length was 512 mm at 1064 nm and 500 mm at 532 nm. By adjusting the beam-wavefront curvatures with the expander/collimators, the beam arrays were allowed to focus on the same plane and to coincide at the same points with the aid of the folding mirror. The intensity distributions of the split beams with ~300 μm spacing are shown in Figs. 11(a) and 11(b), respectively, in comparison with the theoretical predictions in Figs. 11(c) and 11(d). Clearly, the input laser energies were delivered to the target diffraction orders: 0th, ± 1st, and ± 2nd at 1064 nm and 0th, ± 2nd, and ± 4th at 532 nm; and the undesired orders in the array: ± 1st and ± 3rd at 532 nm, were sufficiently suppressed. The efficiencies and uniformities were measured to be 76% and 0.58 at 1064 nm, and 65% and 0.64 at 532 nm. The peaks outside the arrays (the ± 5th, ± 6th, and ± 7th orders at 1064 nm and the ± 7th and ± 10th at 532 nm) were small enough to set an irradiation threshold for laser processing. In this measurement, the beams entered from the diffractive surface and the Fresnel loss was compensated at the rear of the splitter. The optical properties described above were independent of the beam polarizations. In the measurement, the polarizations were adjusted using a commercially available half-wave plate.
Fig. 11 Beam arrays captured by a beam profiler: (a) at 1064 nm and (b) at 532 nm, and theoretical predictions (c) at 1064 nm and (d) at 532 nm.
7. Discussions
The measured performance of the 5-fan-out splitter was not satisfactory, when compared to theoretical predictions, owing to the following errors: (1) uneven resist-layer thickness, (2) resist-surface roughness in the exposed regions, (3) residual nonlinearity in the resist characteristics, and (4) residual nonuniformity in the projection luminance. Among these errors, (3) and (4) are the most problematic, as they recur in the periods illuminated with the laser beams, and therefore the aberrated wavefronts due to these periodic errors interfere constructively, degrading the beam arrays in the focal plane. With regard to tiling errors, a 13 × 10 grid of lines, ~1 μm wide and ~100 nm deep, were formed across the pattern; yet, energy loss due to diffraction by these grooves was less than ~1%. Random-depth errors can be reduced by using only the linear part of the characteristic curve of the resist at higher doses. For this purpose, the resist can be biased to the sensitivity threshold by preexposing it with a uniformly distributed UV light [23, 24]. In this study, however, that requires further thickening of the resist layer to be as thick as ~10 μm, and to improve thickness uniformity, the resist coating recipe must be reexamined. Profile errors caused by luminance variations across the DMD can be eliminated by appropriately mapping them out to modify the pattern data. An NTD (negative tone development) shrink process may be effective for reducing resist roughness [25]. These issues are currently under investigation.
When it comes to applying the splitters to laser materials processing, split counts and beam registration are important issues that must be addressed. The split-count upper bound is limited by the pixelated corrugations, because pixelation affects the intensity uniformity of the beam arrays. This modulation is estimated using sinc2(m/M), where m is the outermost diffraction order in the beam array and M the number of pixels within one period [26]. The period should to be as small as ~100 μm to allow for a short focal length, e.g., 100 mm, to obtain the tightly-focused beam spots needed in laser microprocessing. This target is attainable because, as shown in Fig. 6, a round-off error (pixel/period) of 1% is acceptable provided that the process is accurate and reproducible. To ensure a modulation > 0.99, for instance, the split counts must be 13 or less for M = 100. The wavelength difference is highly relevant to beam registration. Although some applications may use two wavelengths that do not satisfy Eq. (1) exactly, two beam arrays can be obtained and made to overlap precisely at the process points by using a wavefront-adjustment technique, as described in the preceding section. When a beam of wavefront curvature Ri is incident on a lens, the effective focal length is given by Li = (Ri/Ri-fi)fi, with i = 1 or 2, and where fi is the intrinsic focal length at λi and the beam spacing is given by Δ = λiLi/p. The spacing may therefore be tuned to be equal for both wavelengths by adjusting Ri (and hence Li) independently with the beam expander/collimator placed on each beam path. Nevertheless, a small displacement can be introduced between the beam arrays by adjusting the optical configuration, if necessary. As compared to using two conventional DBSs designed at specific wavelengths with each on a separate beam path, the dual-wavelength DBS introduces more flexibility in terms of optical layout. The layout illustrated in Fig. 10, for example, gives a wide tuning range to the beam spacing. This is convenient in laser microprocessing. Furthermore, the proposed splittter saves the labor for aligning the optical parts and the space for the entire laser system when being operated with a dual-emission laser.
The use of the splitters with high power lasers requires that they be fabricated from robust materials, i.e., glass and polymer. A replication process that transfers resist patterns to these materials with a reasonable yield rate needs to be developed. Because the refractive-index dispersion in these materials differs from that in resist, ϕmax and K must be reevaluated using the scheme introduced in Section 3. For example, fused silica has a much smaller dispersion compared to resist.
8. Conclusion
This study demonstrated the feasibility of a dual-wavelength DBS functioning at two different wavelengths, and identified issues requiring further attention. To combine two versatile wavelengths in laser processing, 1064 and 532 nm, we designed surface profiles for splitters using an in-house-developed SA-based design software. We presented a heuristic but effectual scheme for increasing the likelihood of finding the optimized profile. We applied maskless grayscale lithography using a high-resolution DMD to fabricate a 5-fan-out DBS in the photoresist. The diffractive properties of the splitter were evaluated to confirm its dual-wavelength functionality. This achromatic diffractive beam splitter, when applied to laser-based processes, can achieve a high throughput with a simple optical configuration, and will ultimately yield the benefits of dual-wavelength processing.
Acknowledgment
A part of this study was supported by the National Institute of Material Science (NIMS) Nanofabrication Platform. The authors thank Dr. Daiju Tsuya and Dr. Eiichiro Watanabe for their technical support in the fabrication of the splitters and for stimulating discussions.
References and links
1. J. Amako, T. Shimoda, and K. Umetsu, “Versatile light-control schemes based on diffractive optics for laser drilling, cutting, and joining technologies for microelectronic and micromechanical components and devices,” Proc. SPIE 5339, 475–487 (2004). [CrossRef]
2. X. Liu, “Industrial applications of ultrahigh precision short-pulse laser processing,” Proc. SPIE 5713, 372–386 (2005). [CrossRef]
3. T. E. Lizotte and O. P. Ohar, “Fan out diffractive beam splitter: 5 phase level binary versus 50 phase level grayscale,” Proc. SPIE 6290, 62900D1 (2006).
4. O. Haupt, V. Schütz, and U. Stute, “Multi-spot laser processing of crystalline solar cells,” Proc. SPIE 7921, 79210V1 (2011).
5. A. Krasnaberski, Y. Miklyaev, D. Pikhulya, L. Kleinschmidt, W. Imgrunt, M. Ivanenko, and V. Lissotschenko, “Efficient beam splitting with continuous DOEs and microlens arrays,” Proc. SPIE 8236, 823609 (2012). [CrossRef]
6. D. H. Auston, J. A. Golovchenko, and T. N. C. Venkatesan, “Dual-wavelength laser annealing,” Appl. Phys. Lett. 34(9), 558–560 (1979). [CrossRef]
7. J. Zhang, K. Sugioka, T. Takahashi, K. Toyoda, and K. Midorikawa, “Dual-beam ablation of fused silica by multiwavelength excitation process using KrF excimer and F2 lasers,” Appl. Phys., A Mater. Sci. Process. A71, 23–26 (2000).
8. M. Sakamoto, T. Tachikawa, M. Fujitsuka, and T. Majima, “Two-laser-guided three-dimensional microfabrication and processing in a flexible polymer matrix,” Adv. Mater. 20(18), 3427–3432 (2008). [CrossRef]
9. A. Hess, R. Schuster, A. Heider, R. Weber, and T. Graf, “Continuous wave lasers welding of copper with combined beams at wavelengths of 1030 nm and of 515 nm,” Phys. Procedia 12, 88–94 (2011). [CrossRef]
10. S. Noach, A. Lewis, Y. Arieli, and N. Eisenberg, “Integrated diffractive andrefractive elements for spectrum shaping,” Appl. Opt. 35(19), 3635–3639 (1996). [CrossRef] [PubMed]
11. I. Barton, P. Blair, and M. R. Taghizadeh, “Dual-wavelength operation diffractive phase elements for pattern formation,” Opt. Express 1(2), 54–59 (1997). [CrossRef] [PubMed]
12. J. Bengtsson, “Kinoforms designed to produce different fan-out patterns for two wavelengths,” Appl. Opt. 37(11), 2011–2020 (1998). [CrossRef] [PubMed]
13. T. R. M. Sales and D. H. Raguin, “Multiwavelength operation with thin diffractive elements,” Appl. Opt. 38(14), 3012–3018 (1999). [CrossRef] [PubMed]
14. Y. Ogura, N. Shirai, J. Tanida, and Y. Ichioka, “Wavelength-multiplexing diffractive phase elements: design, fabrication, and performance evaluation,” J. Opt. Soc. Am. A 18(5), 1082–1092 (2001). [CrossRef]
15. G. Kim, J. A. Domínguez-Caballero, and R. Menon, “Design and analysis of multi-wavelength diffractive optics,” Opt. Express 20(3), 2814–2823 (2012). [CrossRef] [PubMed]
16. M. N. Vesperinas, F. J. Fuentes, and R. Navarro, “Performance of a simulated-annealing algorithm for phase retrieval,” J. Opt. Soc. Am. A 5(1), 30–38 (1988). [CrossRef]
17. J. W. Goodman, Introduction to Fourier Optics, 3rd ed. (Roberts & Company, 2005).
18. R. Magnusson and T. K. Gaylord, “Diffraction efficiencies of thin phase gratings with arbitrary grating shape,” J. Opt. Soc. Am. A 68(6), 806–809 (1978). [CrossRef]
19. B. Brixner, “Refractive-index interpolation for fused silica,” J. Opt. Soc. Am. 57(5), 674–676 (1967). [CrossRef]
20. http://en.wikipedia.org/wiki/Cauchy_equation
21. http://www.microchemicals.com/micro/AZ%20 P4000%20Phototresist.pdf.
22. http://www.nanosystem-solutions.com/en/product/maskless.html
23. M. LeCompte, X. Gao, and D. W. Prather, “Photoresist characterization and linearization procedure for the gray-scale fabrication of diffractive optical elements,” Appl. Opt. 40(32), 5921–5927 (2001). [CrossRef] [PubMed]
24. H. Andersson, M. Ekberg, S. Hård, S. Jacobsson, M. Larsson, and T. Nilsson, “Single photomask, multilevel kinoforms in quartz and photoresist: manufacture and evaluation,” Appl. Opt. 29(28), 4259–4267 (1990). [CrossRef] [PubMed]
25. M. Shirakawa, N. Inoue, H. Furutani, K. Yamamoto, A. Goto, and M. Fujita, “Advanced patterning approaches based on negative-tone development (NTD) process for further extension of 193nm immersion lithography,” Proc. SPIE 9425, 942509 (2015). [CrossRef]
26. I. M. Barton, P. Blair, and M. R. Taghizadeh, “Diffractive phase elements for pattern formation: phase-encoding geometry considerations,” Appl. Opt. 36(35), 9132–9137 (1997). [CrossRef] [PubMed]
