Fabrication and evaluation of diamond binary diffractive fan-out elements is demonstrated. The diffractive optical elements (DOEs) are designed for two different wavelengths, 633 nm and 10.6 µm. The DOE splits an incident beam into 16 spots that form a ring pattern. The surface reliefs were fabricated by photolithographic methods followed by plasma etching, which produced well-defined patterns with smooth surfaces. One DOE was optically evaluated with a HeNe laser, operating at a wavelength of 633 nm, and showed good performance. The DOE designed for a wavelength at 10.6 µm was tested together with a carbon dioxide laser. The light pattern was used to microstructure a 10 mm thick PMMA piece with very good results.
© 2003 Optical Society of America
High power lasers such as carbon dioxide (CO2) lasers are today important tools for applications in delivering a high amount of energy to small areas. Main markets are material and medical processing. Optical powers, in continuous mode, of up to 10 kW are used for cutting or welding of metal in the industry. This high optical power creates special demands on the output coupler windows of the lasers and the optical components used to shape the laser beams (e.g. focusing or more advanced beam shaping structures). Until today the choice of material has been zinc selenide (ZnSe). This is because ZnSe has a very low absorption coefficient at 10.6 µm wavelength. However, because of the high optical powers used in CO2 lasers, heat will be generated which will create thermal lensing or even crack the ZnSe material. For CO2 lasers of today, ZnSe is already a limiting factor in terms of optical power. Diamond on the other hand, even though it has a higher absorption coefficient than ZnSe at 10.6 µm, has clearly a better potential to withstand these high optical powers [1,2]. This is mainly due to diamond’s high thermal conductivity, which in fact is the highest of all solids known. Diamond has also benefits over ZnSe in terms of extreme resistance against chemicals, its hardness and its resistance against thermal shock. Diamond optical components can also be used in harsh environments and in conditions where durability with respect to water drop impact and solid particle abrasion is needed (e.g. high aerodynamic load). In Table 1 some relevant properties of diamond and ZnSe are shown.
The recent progress in chemical vapor deposition (CVD) diamond technology has opened up the possibility to fabricate diamond optical components on an economically feasible basis. So far the use of diamond has in principle been limited to output coupler windows and little has been done in realizing more complex diamond optics because of the difficulties in machining diamond. Structuring of diamond can be done by laser ablation or focused ion beam milling [3,4], but these are costly and slow technologies, mainly suitable for laboratory demonstrations. Recent progress in micromachining of diamond optical components by highdensity plasma etching shows that this technology has the potential to meet the industry demands on high fidelity of the manufactured structures at low cost . Plasma etching is a well known technology in the microelectronic industry for batch-wise production of semiconductor and metal structures.
This paper demonstrates the fabrication of an advanced diffractive fan-out element fabricated in diamond of optical quality. The diffractive optical element (DOE) was designed for a wavelength at 10.6 µm, which is the operating wavelength for CO2 lasers. The diamond DOE was also optically evaluated with a CO2 laser. Another DOE, based on the same design but optimized for a wavelength of 633 nm, was also fabricated and tested. Using a HeNe laser as the light source allowed us to perform a more thorough optical evaluation than with the CO2 laser.
2. DOE design and fabrication
A detailed explanation of the design procedure for this type of DOEs (fan-out elements) is found elsewhere . In this part we only outline some of the steps.
The phase-modulating surface for the fan-out element was calculated with an optimalrotation angle method. This is a direct Fourier method, which means that no inverse transformation is performed. In this method the height of the binary phase-relief in each discrete point is chosen so that the diffraction efficiency is increased. The designed DOE phase-relief is shown in Fig. 1 and is intended to produce 16 spots which form a ring pattern in the far field.
The intensity is equal in all spots and the zeroth order is not present. When calculating the binary phase pattern, the DOE square was discretized into 128×128 pixels, each pixel being 10×10 µm in size. 128 pixels were chosen since numerical Fourier-transformation algorithms is easiest handled for discretizations of 2n×2n, n being a positive integer. We assumed that the light incident upon the DOE is planar wave fronts (and monochromatic). At a wavelength of 10.6 µm the physical scaling is as follows; every spot in the diffraction plane occupy 0.47° and the diameter of the ring (defined of the 16 spots) is 17.1°. These angles scale linearly with the illuminating wavelength. The ideal depth for binary DOEs is determined by the requirement of a π phase shift. The optimal depth (Do) then becomes (assuming ambient air):
which in our case corresponds to a depth of 3.84 µm and 224 nm for an illuminating wavelength (λo) of 10.6 µm and 633 nm, respectively. The refractive index of diamond, ndiamond, is 2.38 at 10.6 µm and 2.41 at 633 nm. Deviation from this depth will lower the diffraction efficiency of the DOEs, diffracting light into the zeroth order as well as higher orders and unwanted scattering due to the restriction of a binary surface-relief. The difference in phase between the red and the white areas in Fig. 1 is π rad, which means that the white areas are etched to a depth of Do. Calculations showed that a deviation of 10% from the optimal depth resulted in noticeable deterioration of the image. The zeroth order is then present and has an intensity almost comparable with the intensity in the desired spots. The diffraction efficiency is calculated to be 64.0% at 10% deviation from Do, and at a deviation of 16.1% (corresponds to illuminating a DOE designed for red light with green light) the zeroth order will be stronger than the desired spots, and the DOE will have a diffraction efficiency of 61.2%. The deviation will however not affect the intensity distribution between the desired spots, which means that they will all be equally intense/strong regardless of the amount of depth deviation. The maximum theoretically diffraction efficiency (no deviation from the optimal depth) is calculated to be 65.7%. The rest of the light is then diffracted into the higher orders and other unwanted scattered light.
Fabrication of the DOEs was done with contact photolithography, which allows binary elements to be exposed, followed by plasma etching. The DOEs were fabricated in CVD diamond substrates (0.3×Ø10 mm, flatness <2.5 fringe at HeNe). The diamond substrate is polycrystalline CVD diamond grown in microwave plasma, with the first seeding material removed. The surfaces are polished to a root-mean-square (RMS) roughness below 15 nm. Properties of the diamond substrates are shown in Table 1. The material is commercially available from Drukker International B.V., The Netherlands, and is specially made for use as optical transmission components (from visible light to the far infrared).
Since we wanted to structure the whole substrate, the binary phase pattern was duplicated side by side in the CAD software to an area of 10×10 mm, the pattern was then transfered to an chromium mask with standard laser beam writing. When the whole substrate is structured no demands of an exact positioning of the laser beam on the diamond DOE is needed (in the absence of a focusing lens) which makes the optical evaluation easier, especially with the CO2 laser.
The diamond substrates were first cleaned in warm H2SO4 and H202 (equal parts). Then a cleaning in warm NH3, H20 and H202 and a clean in HCl, H2O and H2O2 and finally a dip in HF and water was performed. After the cleaning steps the diamond substrate was sputtered with a 200 nm thick aluminum layer. In next step a thin layer of photoresist (Shipley 1813) was spun out on top of the aluminum. Since the substrates are quite small it is favorable to spin as thin layer as possible (e.g. high speed) to reduce the edge effects of the resist. At slower rotation speed the edge thickness will be larger and therefore reduce the possibilities to get small feature size of the resist structures (due to air gap in the middle of the substrate between the chromium mask and the resist). After a soft bake on a hot plate (1 min., 115°C), the DOE pattern was transfered to the photoresist with contact lithography followed by resist development and hard bake of the remaining resist (hotplate: 1 min., 115°C). The DOE pattern in resist was transfered into the aluminum by wet or dry etching of aluminum. Wet etching is done in a standard Al-etch (based on H3PO4) and in the case of dry etching Cl2 and BCl3 was used as etch gases in a commercial inductively coupled plasma (ICP) etching system. With this types of structures, with rather large size (~10µm), wet etching works well but in the case of smaller structures dry etching is preferred.
3. Plasma etching of the diamond DOE
To fabricate the binary phase pattern in diamond a home built ICP etching system was employed. In earlier research we have showed that diamond etching in ICP etching system generates high etch rates and smooth pattern transfer . The etch gases used was O2 and Ar with gas flows of 7 sccm and 8 sccm, respectively. ICP power was 600 W, bias -140 V and chamber pressure 2.5 mTorr. All samples were mounted with vacuum grease on the water cooled aluminum rf-chuck to enhance the thermal conductivity (eg. avoid burning of the resist). The Al-pattern on the diamond substrate served as etch stop during the diamond etching. By first measuring the etch rate of partly covered diamond and knowing the desired relief depth, we could easily calculate the etch time needed for fabricating the DOEs. The etch rate of diamond was measured to be 200 nm/min which corresponds to a total etch time of 19 min for λ0=10.6 µm and 67 s for λo=633 nm. Finally the Al was stripped by wet etching. The etch depth and surface roughness was measured with a white light interferometer. The rms roughness was estimated to be 5–15 nm, which is in the same order as the unetched diamond surface. In Fig. 2 a SEM picture of diamond DOE which is etched to a depth of 3.84 µm (for use with CO2 laser) is shown. Another substrate was etched to depth of 224 nm, designed for use with a He-Ne laser operating at 633 nm. These two DOEs were then optically characterized.
4. Optical characterization
The two samples, one designed for visible light and the other for the IR region, was characterized by illuminating the fabricated DOEs with a laser at the design wavelengths. The resulting intensity distributions in the far field were then evaluated.
4.1. He-Ne laser (λ=632.8 nm)
The diamond fan-out element was evaluated with a 17 mW He-Ne laser operating at 632.8 nm (red light). We also illuminated the DOE with a 1 mW He-Ne laser operating at 543.5 nm (green light) to see how sensitive the DOE is to depth errors and compare with calculated values. Using green light corresponds to a depth error of 16.1%. The setup is simple in this experiment, the diamond DOE was placed in the laser beam and the intensity pattern generated in the far field was recorded with a CCD-camera. The intensity in each spot (and the total amount of light) was measured with a power and a radiance meter. In Fig. 3 the recorded diffraction pattern for red light and green light is showed. Measurement of the intensity distribution compares well with the calculated. The measured diffraction efficiency was 64.5% for red light and 57% for green light. These measured values should be compared with the calculated values, 65.7% for red light and 61.2% for green light. The left hand side of Fig. 3 shows the diffraction pattern at λ=633 nm, where we have used a filter in front of the CCD-camera so as not to saturate the detector. Although it cannot be seen in the left side of Fig. 3, a faint zeroth order is present. By evaluating the measurement data this zeroth order was found to amount to 5% of each desired diffraction spot.
A commonly chosen quality measure used to characterize a fan-out element is the uniformity error (U) defined as:
where Imax and Imin is the power of the brightest and faintest spot, respectively. The spots forming the ring pattern have all similar intensities, which indicates that the fabrication process of the DOE has been accurate. Uniformity error, using red light, was measured to be 4.0%. The small uniformity error can probably be attributed to slightly varying depth over the DOE structure as well as some rounding of the binary phase-relief edges. As expected, the zeroth order is present when green light is used (see Fig. 3). The zeroth order was measured to be 59% stronger than each of the desired spots, which is in good agreement with the theoretical value of 62%. The uniformity error is measured to 4.4%, this is expected since depth deviation have influence on the diffraction efficiency but not on the uniformity error (for this type of binary DOE). The angle of the ring pattern was measured for red and green light to 1.015° and 0.870° respectively. Calculated values are 1.020° and 0.876° for red and green light.
4.2 CO2 laser (λ=10.6 µm)
In this part we evaluated the diamond DOE with a CO2 laser operating in cw mode at a wavelength of 10.6 µm. The system is equipped with a time controlled shutter system to block the laser beam. Figure 4 shows the experimental setup. The diamond substrate was placed in a brass holder which was in contact with a cold water system. The beam width of the CO2 laser beam impinging on the diamond DOE was approximately 8 mm. The beam emitted from the laser is slightly diverging and contains no other modes except the fundamental TEM00-mode. The focal distance in this setup is 127 mm. To record the intensity pattern from the DOE we used a 10 mm thick Poly(methylmethacrylate) (PMMA) plate, placed at the focal plane. By this way the laser beams “drilled” holes in the PMMA plate.
Figure 5 shows a PMMA plate microstructured by four different exposures of the CO2/diamond DOE combination for 0.5 s. The optical power was 600 W which resulted in an approximate optical intensity of 800 W/cm2 at the DOE. The diffraction efficiency could not be calculated since no power meter that can handle this high power with good lateral resolution was available. However, one can see that uniformity is very good by measuring the depth of the holes drilled in PMMA, which all have equal depths (9 mm) and diameter. A small fraction of the light could be found in the zeroth order. Also, four spots placed in a symmetrical way around the zeroth order (see Fig. 5) were found. By a more thorough investigation of the Fourier transformed phase pattern we noticed that this in fact was relating to the DOE design. The power in each of these spots was approximately equal to the zeroth order. The angle occupying of the ring pattern was 18.1°, which is compared with the calculated value of 17.1°.
5. Discussion and Conclusions
We have fabricated and optically evaluated a diamond fan-out DOE, designed for use with CO2 lasers. A 1-cm thick PMMA plate was successfully microstructured in this way. A diamond DOE has also been designed for red light. The evaluated DOEs showed almost perfect performance compared with calculated values.
We have demonstrated a process for fabricating DOEs in diamond which can open new possibilities in the high power laser industry in terms of much longer life-time of the optical elements (compared to ZnSe) and ability to use higher optical power. We now intend to fabricate continuous relief DOEs for CO2 lasers by replication in suitable resist followed by plasma etching in diamond. With suitable resist we mean that the resist should have as low an etch rate (in the diamond plasma etch process) as possible, for instance SU-8 or Benzocyclobutene (BCB) which typically have slow etch rates in oxygen plasma (compared to Shipley- and similar positive resists). Furthermore, continuous-relief elements typically allow a higher diffraction efficiency than binary phase-reliefs.
The authors gratefully acknowledge Hans Engström and Klas Nilsson at the Division of Manufacturing Systems Engineering, Luleå University of Technology for help with the CO2 laser experiments and Jörgen Bengtsson at Chalmers University of Technology for help with the design of the DOE. This work was in part financed by SUMMIT, the Swedish Center for Surface and Microstructure Technology, supported by the Swedish Agency for Innovation Systems (VINNOVA).
References and links
1. C. J. Brierley, C. M. Beck, G. R. Kennedy, J. Metcalfe, and D. Wheatley, “The potential of CVD diamond as a replacement for ZnSe in CO/sub 2/ laser optics,” Diamond and Related Mater. 8, 8–9 (1999). [CrossRef]
2. B. Dischler and C. Wild, Low-Pressure Synthetic Diamond (Springer, Berlin Heidelberg, 1998). [CrossRef]
3. S. Gloor, V. Romano, W. Luthy, H. P. Weber, V. V. Kononenko, S. M. Pimenov, V. I. Konov, and A. V. Khomich, “Antireflection structures written by excimer laser on CVD diamond,” Appl. Phys. A 70, 547–550 (2000). [CrossRef]
4. J. D. Hunn and C. P. Christensen, “Ion beam and laser-assisted micromachining of single crystal diamond,” Solid State Technol. 37, 57–60 (1994).
5. M. Karlsson, K. Hjort, and F. Nikolajeff, “Transfer of continuous-relief diffractive structures into diamond by use of inductively coupled plasma dry etching,” Opt. Lett. 26, 1752–1754 (2001). [CrossRef]
6. J. Bengtsson, “Design of fan-out kinoforms in the entire scalar diffraction regime with an optimal-rotationangle method,” Appl. Opt. 36, 8435–8444 (1997). [CrossRef]