1. Introduction

Two-photon polymerization based on scanning irradiation of femtosecond laser pulses is an interesting tool for fabrication of microstructures [1–6]. The technique has been applied to three-dimensional (3D) microstructures [2,3,5,6]. Recently, sub-wavelength writing has been demonstrated by some groups [7–9]. Because such 3D microstructures have numerous irradiation points with a sub-wavelength unit size, a femtosecond laser irradiation system with high throughput and high energy utilization efficiency will be required to fabricate devices with a practically effective size. Parallel femtosecond laser irradiation techniques can satisfy these demands [10–12]. Additionally, parallel femtosecond laser irradiation using a computer-generated hologram (CGH) displayed on a spatial light modulator (SLM) is a powerful technique, because a CGH can generate a desired arbitrarily variable beam, such as a beam with a prescribed spatial shape, a split beam, a focused beam, or a wavefront corrected beam, with low loss of light [13–15]. This technique, which is called holographic femtosecond laser processing, can also be directly applied to two-photon polymerization with the above advantages. Recently, parallel two-photon polymerization with a hologram has been demonstrated [16].

In the study described in this paper, we demonstrated a sparse exposure technique combining parallel exposure and scanning exposure. This technique improved the fabrication throughput and enabled simultaneous fabrication of linear structures with different widths. We also demonstrated the fabrication of space-variant structures by changing a CGH, as well as parallel fabrication of voxel structures with single femtosecond laser pulse irradiation.

2. Experimental setup

Figure 1 shows the experimental setup we used for holographic two-photon polymerization. The system mainly consisted of an amplified femtosecond laser system, a liquid-crystal spatial light modulator (LCSLM; Hamamatsu Photonics, PPM) [17], and relay optics. The amplified femtosecond laser system consisted of a mode-locked Ti:sapphire laser (Spectra Physics, Tsunami) pumped by a diode-pumped, solid-state continuous-wave green laser (Spectra Physics, Millenia), and a pulsed Ti:sapphire regenerative amplifier (Spectra Physics, Spitfire) pumped by a diode-pumped, Q-switched Nd:YLF laser (Spectra Physics, Merlin). The femtosecond laser system generated pulses with a peak wavelength of λ_p=800 nm and a width of ~150 fs. The laser pulse, collimated by a beam expander, was diffracted by a CGH displayed on the LCSLM to form a fabrication pattern at the Fourier plane P of Lens 3. The 0-th order beam was obstructed at the plane P. The fabrication pattern was applied to the sample by reduction optics composed of Lens 4 and a 100× microscope objective lens (OL) with a numerical aperture (NA) of 1.25. The irradiation energy at the sample was given by the product of the energy measured before introducing the laser pulse into the OL in each experiment and the OL’s transmittance of 0.5 that was measured in advance. The axial and lateral processing positions of the sample were controlled by a three-dimensional motorized stage. A halogen lamp (HL) and a charge-coupled device (CCD) image sensor were used to observe the processing.

 

Fig. 1. Experimental setup.

Download Full Size | PDF

The size of a pixel on the sample plane in the experimental setup, corresponding to that in the computer, was calculated. The CGH with 512×512 pixels was designed with the optimum rotation angle (ORA) method [18], with compensation for the spatial frequency response of the LCSLM [19,20], and was displayed on a W ²=13×13 mm² area on the LCSLM. The focal length of the first Fourier transform lens L3 was F ₃=600 mm, the position x of the diffraction beam from the grating with the maximum spatial frequency ν _max of 19.2 line pairs/mm (lp/mm) was 9.22 mm, as calculated from x=λ_p F ₃ ν _max. Therefore, the size of a pixel on the plane P was 36 µm. Because the focal lengths of Lens 4 and OL were F ₄=200 mm and F _OL=1.9 mm, respectively, the pixel size on the sample plane was geometrically corresponding to p _s=342 nm. The pixel size was an important system parameter because it decided the position resolution of fabricated structures. The beam radius at the pupil of the OL was 2.17 mm (=W/2×F ₄/F ₃), and the effective NA calculated from this beam radius was NA_eff=1.14.

The sample was spin-coated negative photoresist (SU-8 3000, Microchem) with a film thickness of 7 µm. A scanning electron microscope (SEM; Hitachi, S4700) was used to observe the structures.

3. Experimental results

Figure 2 shows the experimental result of simultaneous microfabrication of line structures with different widths using the sparse-exposure technique. Figure 2(a) shows a CGH with 10 diffraction peaks. The minimum and maximum spatial frequencies were 1.59 and 2.99 lp/mm, respectively. Figure 2(b) shows its optical reconstruction observed on the plane P. The peak at the left in the picture is the 0th-order beam, which was obstructed when exposure was performed. Figure 2(c) shows an SEM image of the fabricated linear structure when the sample was exposed using the CGH, with lateral translation of the sample stage in the direction indicated by the arrow. The repetition of the laser pulses was 1kHz and the irradiation energy was 40 nJ/pulse. The scanning speed of the stage was 100 µm/s. The widths of the fabricated linear structures were 1.9 µm, 0.78 µm, and 3.6 µm from the top of the picture, respectively.

In this demonstration, the important technique was that the diffraction peaks were positioned sparsely to avoid interference between them. The distance between the diffraction beams should be sufficiently longer than the beam diameter. The pixel size on the sample plane was p _s=342 nm, and the beam diameter was d _Airy=856 nm, in terms of the Airy disk size calculated from NAeff and λ_p. In order to avoid overlapping of the central Airy disk, it is required that the adjacent diffraction peaks be spaced by more than 2.5 pixels (=d _Airy/p _s). The distance between the nearest peaks of the diffraction shown in Fig. 2(b) was 6 pixels; therefore, the diffraction peaks did not interfere. The number of pixels in the vertical direction of the sample motion, indicated by x, was 2. Because 2p _s<d _Airy, the linear structures exposed by the diffraction peaks joined the thick linear structures. The sparse exposure technique thus enabled simultaneous fabrication of linear structures with different widths.

 

Fig. 2. Sparse exposure technique. (a) CGH, (b) its optical reconstruction of the CGH, and (c) SEM image of three linear structures with different widths.

Download Full Size | PDF

The width of the linear structure depends on the scanning speed and the energy of the diffraction peaks. The dependency was examined by using the CGH with ten diffraction peaks shown in Fig. 3(a). The minimum and maximum spatial frequencies of the diffraction peaks were 1.59 lp/mm and 4.96 lp/mm, respectively. Figure 3(b) shows the computer reconstruction and its intensity profile, and Fig. 3(c) shows the optical reconstruction and its intensity profile on the plane P. In the computer reconstruction, the uniformity U was 99.86%, and the diffraction efficiency η was 69.2%. In the optical reconstruction, U was 91% and η was 61%. Here, the uniformity is defined as U=I _min/I _max, where I _min and I _max are the minimum and the maximum peak intensities, respectively. The diffraction efficiency η is defined as the ratio of the 1st-order diffracted light intensity to the 0th-order light intensity when the LCSLM does not display a CGH. The degradation of U from the computer to the experimental reconstructions was caused by the incompleteness of the compensation for the spatial frequency response of the LCSLM in the computer [19,20].

Figures 4 show the optical reconstruction of a CGH used for fabricating supports on the plane P. The supports were applied to measure the size of the linear structures fabricated by the CGH by detaching them from a glass substrate. The use of the supports prevented to lose the fabricated structures during development. This technique was first demonstrated by Seet et al. [21]. For easy understanding, the center peak of the 0th-order beam was not obstructed when the picture in Fig. 4 was taken. Each of the thick supports was fabricated by the sparse exposure technique with scanning of the three diffraction beams.

 

Fig. 3. (a). CGH for generating ten beams, (b) its computer reconstruction, and (c) its optical reconstruction.

Download Full Size | PDF

 

Fig. 4. Optical reconstruction of the CGH for fabrication of supports. The center peak is the 0-th order beam.

Download Full Size | PDF

Figure 5 shows an SEM image of the ten linear structures with the four thick linear supports fabricated by the holographic two-photon polymerization. The ten linear structures were fabricated with a pulse repetition of 1kHz, an irradiation energy of 40 nJ/pulse, and a scanning speed of 100 µm/s. The supports were fabricated with an irradiation energy of 110 nJ/pulse and a scanning speed of 25 µm/s. They were separated from the glass substrate and supported with four thick linear structures so as not to be lost during development. The adjacent linear structures were contacted with the outsides of the supports in a drying process after the development [22].

Figure 5(c) shows the width of the linear structure for an average irradiation energy of ten diffraction peaks. The filled circles, the filled triangles, and the rectangles indicate the widths of the linear structures fabricated when the stage speeds were 6.24 µm/s, 25 µm/s, and 100 µm/s, respectively. The solid curves were calculated by the equation d=w[ln(E/E _th)]1/2, where w (=d _Airy) is the beam diameter, E is the irradiation energy, and E _th is the threshold energy for fabrication [21]. The values of E _th were estimated as 0.8, 1.3, and 2.0 nJ/pulse, respectively. The minimum linear structure was 230 nm under an average irradiation energy of 1.0 nJ/pulse and a scanning speed of 6.24 µm/s, in our present experimental setup. When the scanning speed was low and the irradiation energy was high, the dependency of the width on the irradiation energy was larger than the value expected by the equation because of a micro explosion induced by the laser pulse. At the scanning speed of 6.24 µm/s, when the average irradiation energy was more than 3.0 nJ/pulse, a distorted complex structure was fabricated, and its width could not be measured.

 

Fig. 5. SEM images of ten linear structures with four thick linear supports fabricated by holographic two-photon polymerization. (a) Top and (b) oblique views. (c) The width of the linear structures for the average irradiation energy of ten diffraction peaks. The scanning speeds of the sample stage were 6.24 µm/s (circles), 25 µm/s (triangles), and 100 µm/s (squares), respectively.

Download Full Size | PDF

 

Fig. 6. SEM images of two-dimensional structure formed by switching the 15 CGHs. (a) Top view and (b) oblique view.

Download Full Size | PDF

Figure 6 is a demonstration of the holographic two-photon polymerization. Fifteen CGHs calculated in advance were switched at 2 Hz with a stage speed of 6.24 µm/s. When the CGH with 20 diffraction peaks, which was the maximum number in this experiment, was used, the pulse repetition was 1kHz and the irradiation energy was 70 nJ/pulse. When the number of diffraction peaks was small, the CGH was designed to ensure that the irradiation energy was 3.5 nJ/pulse per linear structure. The width of each structure was about 700 nm.

Figure 7 show the first demonstration of a voxel structure fabricated with single femtosecond laser pulse irradiation. The CGH generated ten diffraction peaks. The total irradiation energy was 150 nJ. The structures were fabricated while changing the axial focus position in the glass substrate. Figure 7(a) shows an oblique SEM image of the fabricated structures. When the focus was far from the substrate, the contact area between the fabricated voxel structure and the glass substrate became small, consequently, the fabricated voxel structure was pushed down and was washed out during development. On the other hand, when the focus was toward the inside of the glass substrate, the fabricated structure became small and eventually disappeared. Figure 7(b) shows 10 voxels with a diameter of ~400 nm, indicated by the arrow in Fig. 7(a). Figure 7(c) shows the voxel diameter for the average irradiation energy of ten diffraction peaks. Figure 7(d) shows the fabricated voxel when the irradiation energy was 10 nJ. The diameter was measured as shown in the figure. When the irradiation energy was more than 25 nJ, the fabricated structures were distorted by formation of a micro bubble, as shown in Fig. 7(e). The range of the irradiation energy for the fabrication of voxels was narrow, at only 10 to 25 nJ. As indicated by the solid curve, E _th was estimated to be 8.0 nJ.

4. Conclusion

We demonstrated holographic two-photon polymerization using a sparse exposure technique based on a combination of parallel exposure and scanning exposure. The sparse exposure technique enabled the simultaneous fabrication of linear structures with different widths. The technique is expected to improve the throughput in large-scale fabrication. We also demonstrated fabrication of space-variant structures by changing a CGH, which is a major advantage of holographic two-photon polymerization. We further demonstrated parallel fabrication of voxel structures with single femtosecond laser pulse irradiation.

 

Fig. 7. (a). SEM image of ten voxels fabricated by a single laser pulse. (b). Close-up view of the region indicated by the arrow in (a). (c). The voxel diameter for the average irradiation energy of ten diffraction peaks. Close-up views of the voxels when E=10 nJ and 25 nJ, respectively.

Download Full Size | PDF