1. Introduction

Femtosecond laser processing is an important tool for fabricating three-dimensional optical devices in transparent materials [1–5]. In order to fabricate such optical devices at millimeter-scale, an enormous number of processing points are required, and therefore, the processing throughput must be improved. To overcome this issue, parallel femtosecond laser processing based on an array of spots has been proposed. Several methods have been used to demonstrate parallel laser processing, for example, methods using multi-beam interference, microlens arrays, and diffractive optical elements (DOEs). A computer-generated hologram (CGH) [6] allows arbitrary control of the spatial pulse shape, and a spatial light modulator (SLM) displaying a dynamic CGH has been used to achieve variable spatial shaping of femtosecond pulses [7,8]. Femtosecond laser processing with a CGH, called holographic femtosecond laser processing [9–22], has the advantages of high-throughput pulsed irradiation and high energy-use efficiency of the pulse and has been widely used in many applications, for example, two-photon polymerization [23–28], optical waveguide fabrication [29–31], fabrication of volume phase gratings in polymers [32], surface structuring of silicon [33], and cell transfection [34].

In the holographic femtosecond laser processing system, the processing throughput is expressed as the product of the maximum frequency of the SLM, f_SLM, and number of parallel pulses, P, produced by the CGH. The maximum P depends on the total number of pixels of the SLM. Actually, however, the interval between the diffraction spots has to be sufficiently larger than the Airy disk diameter of the spots to avoid distortion of the spots caused by mutual interference between their side lobes. We empirically knew that the minimum interval is about three-times larger than the full width at half-maximum (FWHM) of the spot diameter d_Airy, according to our previous work [18]. The FWHM is approximately equal to the pixel size on the reconstruction plane. This means that mutual interference prevents effective use of the SLM.

In this study, we propose a novel optical system to solve this issue. The system is based on arbitrary polarization distribution control in parallel femtosecond pulses using a pair of SLMs. Although there have been reports of applying polarization control to single beam [35–41], there have not yet been any reports on polarization control of parallel beams. With this system, the interval between the diffraction spots was the closest by avoiding mutual interference. The interval was reduced to half compared with our previous work [18]. We also demonstrated the parallel fabrication of periodic nanostructures [42] with orientation control, which, to our knowledge, is the first reported demonstration of its kind.

2. Principle of polarization control in parallel beams

The optical setup used for polarization distribution control of parallel beams mainly consists of a half wave plate (HWP), a quarter wave plate (QWP), and a pair of SLMs. When the Jones matrixes of these optical elements are defined as H, Q, and S, respectively, the Jones vector of the output beam, E_out, is expressed as

Figure 1(a) shows the direction of linear polarization of the output beam versus phase β. The direction is linearly rotated with a slope of -β/2 by adjusting the phase β. Figure 1(b) shows an example of polarization distribution control using a combination of α and β in Eq. (1). The phase maps for α and β mean the phase distribution on the SLMs, respectively. The gray scale indicates the phase delay. In the input beam with the linear polarization, a component parallel to an axis of SLM receives the phase delay. When the radially polarized beam shown in case 1 in Fig. 1(b) is tightly focused by an objective lens having a high numerical aperture (NA), the vector sum of each polarization component generates a longitudinal electric field parallel to the optical axis at the center of the focal point. This polarization state, called z-polarization, has been shown to enable selective excitation of anisotropic molecules [35], observation of 3D orientation of a single molecule [36], and focusing to a small spot size beyond the diffraction limit [37]. Furthermore, in the focused irradiation at a sample using the radially and azimuthally polarized beam shown in case 2, each polarization component at the focal point behaves as a pure p- or s-polarization, respectively. The reflectance of light with the s-polarization is usually larger than that with the p-polarization at a large incident angle. In particular, the reflectance drastically differs in laser processing of metal materials. Therefore, the radially polarized beam is useful for effective laser cutting, because the beam interacts with the material while constantly maintaining the p-polarization regardless of the beam sweep direction [38]. The azimuthally polarized beam, on the other hand, is useful for effective laser drilling [39], because the wall of the hole is highly reflective to the s-polarized light and does not absorb the incident beam, allowing most of the power to reach the bottom of the hole. There has also been a report regarding improved trapping efficiency of optical tweezers because the reflectance of the particle to the incident light is decreased [40]. Our proposed system also enables these unique properties in parallel beams.

 

Fig. 1 (a) Direction of linear polarization of output beam versus phase β. (b) Typical output polarization distributions versus input signals α and β to the SLMs. The gray scale indicates the phase delay (black and white indicate 0 and 2π, respectively).

Download Full Size | PDF

3. Experimental setup

Figure 2(a) shows a parallel femtosecond laser irradiation system with arbitrary polarization distribution control. It is mainly composed of an amplified femtosecond laser system (Coherent, Micra and Legend Elite), liquid-crystal-on-silicon SLMs (LCOS-SLM; Hamamatsu Photonics, X10468-07), laser processing optics, and a personal computer (PC). The incident femtosecond pulse had a center wavelength of 800 nm, a spectral width of 8 nm FWHM, a pulse width of 110 fs, a repetition frequency of 1 kHz, and linear polarization with a p-component. The pulse was radiated onto the first SLM (SLM1) which displayed CGH1 for applying a pure phase delay to the p-component. The HWP was arranged at a rotation angle of π/8 to rotate the linear polarization by π/4. The pulse was also radiated onto the second SLM (SLM2) which displayed CGH2 for applying a phase delay between the p- and s-components. SLM2 was located at the image plane of SLM1. The circular or elliptical polarization reflected from SLM2 was converted to linear polarization using a QWP set at a rotation angle of π/4. The direction of the linear polarization depended on the phase signal applied to SLM2, as shown in Fig. 1(a). Consequently, the reconstructed spot array with a desired intensity and linear polarization distribution was obtained at plane P and was directed to the laser processing optics, containing a 60 × objective lens (OL) with NA = 0.85. To observe laser processing of the sample, the sample was illuminated with a halogen lamp (HL), and a charge coupled device (CCD) image sensor captured images of the sample via a dichroic mirror (DM) and an infrared (IR) cut filter. The sample was an indium tin oxide (ITO) film (thickness of 10 nm) on a glass substrate. In this holographic setup, an optical transmittance from the femtosecond light source to a Fourier plane of SLM1, a Fourier plane of SLM2, a front of the objective and a behind the objective were 24.7, 11.3, 8.0 and 3.7 %, respectively. Therefore, the transmittance of the whole setup was quite low compared with the ordinary setup without the SLMs. In the case of the sub-micro machining with the femtosecond pulse, however, the holographic setup become predominant than the ordinary setup in terms of the energy-use efficiency of the pulse, because there is a large gap between a threshold energy required to process the sample and a total energy owned by the femtosecond light source. Figures 2(b) and 2(c) show CGHs, designed with the optimal-rotation-angle (ORA) method [43], displayed on SLM1 and SLM2, respectively.

 

Fig. 2 (a) Parallel femtosecond laser irradiation system with arbitrary polarization distribution control. (b), (c) CGHs displayed on SLM 1 and SLM2, respectively.

Download Full Size | PDF

4. Experimental results

Figure 3 shows reconstructed spot arrays using the CGHs shown in Fig. 2(b) and 2(c). The reconstructed spot arrays were captured using the cooled CCD image sensor at plane P in Fig. 2(a). We analyzed the intensity distribution of the reconstructed spot array versus the interval d between the diffraction spots, defined as the distance between the peak intensities. At plane P, the FWHM of the Airy disk pattern in the diffraction spot, d_Airy, was theoretically calculated to be 68 μm from d_Airy = 0.51 × λ / NA. Figure 3(a)-3(d) show the reconstructed spot arrays of the CGH designed with d = 2.88 × d_Airy. The arrows indicate the polarization direction of each diffraction spot. Figure 3(a) shows the optical reconstruction of a parallel spot array without polarization distribution control. A slight distortion in the intensity profile of the diffraction spots was observed due to mutual interference between neighboring spots. Figure 3(b) shows a computational reconstruction without the polarization control. Figure 3(c) shows an optical reconstruction with polarization control. No distortion in the intensity profile was observed because neighboring spots were designed to have mutually orthogonal polarizations. Figure 3(d) shows the reconstructed spot array in Fig. 3(c) captured through a linear polarizer set at two orthogonal angles. It was confirmed that the neighboring spots had mutually orthogonal polarizations. Figure 3(e) shows a spatial frequency analysis of the intensity profiles in Fig. 3(a)-3(c). In Fig. 3(e), the filled circles, solid line, and open circles indicate the spatial frequency spectrum, I(ν), of the intensity profiles in Fig. 3(a), 3(b) and 3(c), respectively. On the horizontal axis in Fig. 3(e), the maximum spatial frequency, ν_μαξ = 60.2 lp/mm, depends on the pixel size of the cooled CCD image sensor, P_s = 8.3 μm. I(ν) of the optical reconstruction without polarization distribution control (filled circles) agreed well with that of the computational reconstruction without polarization control (solid line). Above ν = 25 lp/mm (d = 0.59 × d_Airy), it seems that noise from the CCD image sensor dominated because there was no value in the result for the computational reconstruction (solid line). At ν = 15-25 lp/mm (d = 0.59 × d_Airy to 0.98 × d_Airy), I(ν) with polarization control (open circles) was decreased compared with that without polarization control (filled circles and solid line). It was assumed that ν = 15-25 lp/mm corresponds to the frequency derived from the mutualinterference between diffraction spots, with dependence on the interval d and NA of the focal spot.

 

Fig. 3 (a) Optical and (b) computational reconstructions without polarization control. (c) Optical reconstruction with polarization control and (d) the reconstructed spot array captured through a polarizer set at two orthogonal angles. The interval between diffraction spots, d, was set to 2.88 × d_Airy. (e) Spatial frequency spectrum I(ν) of intensity profiles in the optical (filled circles) and computational (solid line) reconstructions without polarization control and in the optical reconstruction with polarization control (open circles).

Download Full Size | PDF

Figure 4 shows the reconstruction of the CGH designed with d = 1.44 × d_Airy. Figure 4(a) shows the optical reconstruction without polarization distribution control. Considerable distortion in the intensity profile was observed due to mutual interference between neighboring spots. Figure 4(b) shows the computational reconstruction without polarization control. Figure 4(c) shows the optical reconstruction with polarization control. There was some variance in the peak intensities due to slight mutual interference. A smooth intensity profile, however, was maintained. Figure 4(d) shows the reconstruction in Fig. 4(c) captured through a linear polarizer set at two orthogonal angles. In Fig. 4(d), slight distortion was observed due to the mutual interference between spots at d = 2.88 × d_Airy. Figure 4(e) shows a spatial frequency analysis of the intensity profiles in Fig. 4(a)-4(c). In Fig. 4(e), the filled circles, solid line, and open circles indicate I(ν) of the intensity profiles in Fig. 4(a), 4(b) and 4(c), respectively. At ν = 15-25 lp/mm (d = 0.59 × d_Airy to 0.98 × d_Airy), I(ν) of the result with polarization control (open circles) was significantly decreased compared with that without polarization control (filled circles and solid line), like the result of Fig. 3(e).

 

Fig. 4 (a) Optical and (b) computational reconstructions without polarization control. (c) Optical reconstruction with polarization control and (d) the reconstructed spot array captured through a polarizer. The spot interval d was set to 1.44 × d_Airy. (e) Spatial frequency spectrum I(ν) of intensity profiles in the optical (filled circles) and computational (solid line) reconstruction without polarization control and in the optical reconstruction with polarization control (open circles).

Download Full Size | PDF

We quantitatively analyzed the influence of the mutual interference between the diffraction spots by comparing the sum of I(ν) versus the interval d (Fig. 5). In the calculation of the sum of I(ν), I(0) was removed. In Fig. 5, the filled and open circles indicate the results of the optical reconstruction without and with polarization control, respectively. In the result without polarization control (filled circles), the influence of the mutual interference began to appear at d = 2.88 × d_Airy, because the sum of I(ν) was larger than that with polarization control (open circles). This result agreed well with our previous research [18]. In the result with polarization control (open circles), on the other hand, there was no remarkable influence of the mutual interference at d = 1.72 × d_Airy. The sum of I(ν) increased at d = 1.44 × d_Airy due to the slight distortion of the intensity profile, as shown in Fig. 4(c); however, a smooth intensity profile was maintained. As shown in the inset of Fig. 5, it was difficult to avoid the mutual interference at d = 1.15 × d_Airy. By employing the polarization distribution control ofthe parallel beams, the interval d was reduced by half (1.44 × d_Airy / 2.88 × d_Airy) compared with our previous work [18].

 

Fig. 5 Sum of spatial frequency spectrum I(ν) in the intensity profile versus interval d between diffraction spots. The filled circles and open circles are the results without and with polarization control, respectively.

Download Full Size | PDF

Finally, laser processing of an ITO film using parallel beams with arbitrary polarization distribution control was performed. Figure 6(a) shows the reconstruction with polarization distribution control. The interval d was set to d = 5.76 × d_Airy. Figure 6(b) shows a scanning electron microscope (SEM) image of nano-gratings processed using the reconstructed spot array. The processing was performed while irradiating the sample with the reconstructed spot array using a fluence of 1.26 μJ, and scanning the sample at a speed of 20 μm/s. The fluence means the single pulse fluence per spot. From the SEM image, the structures were aligned perpendicularly to the polarization direction of each diffraction beam and had a grating pitch of 140 nm. This is the first demonstration of parallel femtosecond laser processing with arbitrary polarization distribution control.

 

Fig. 6 (a) Optical reconstruction with polarization control. (b) SEM images of structures processed by scanning the reconstructed spot array on an ITO film on a glass substrate.

Download Full Size | PDF

5. Conclusion

We have proposed a parallel femtosecond pulse irradiation system with arbitrary polarization distribution control using a pair of SLMs. The proposed system is based on the simultaneous control of the wavefront and polarization of a femtosecond pulse. By using the system, the interval between the diffraction spots was the closest by avoiding mutual interference. The closest interval was improved to half compared with our previous work. We also demonstrated parallel fabrication of periodic nanostructures with orientation control, which, to our knowledge, is the first demonstration of its kind. Our proposed system will be useful for not only high-speed fabrication of periodic nanostructures with femtosecond laser processing systems, but also imaging of the three-dimensional (3D) orientation of molecules in nonlinear optical microscope systems.