Writing waveguides inside monolithic crystalline ... - OSA Publishing

2 downloads 0 Views 2MB Size Report
Writing waveguides inside monolithic crystalline silicon with nanosecond laser pulses. M. CHAMBONNEAU,1,* Q. LI,1 M. CHANAL,1 N. SANNER,1. AND D.
Letter

Vol. 41, No. 21 / November 1 2016 / Optics Letters

4875

Writing waveguides inside monolithic crystalline silicon with nanosecond laser pulses M. CHAMBONNEAU,1,* Q. LI,1 M. CHANAL,1 N. SANNER,1

AND

D. GROJO1,2

1

Aix-Marseille University, CNRS, LP3, F-13288 Marseille, France e-mail: [email protected]‑mrs.fr *Corresponding author: [email protected]‑mrs.fr

2

Received 9 August 2016; accepted 13 September 2016; posted 19 September 2016 (Doc. ID 273388); published 19 October 2016

Direct three-dimensional (3D) laser writing of waveguides is highly advanced in a wide range of bandgap materials, but has no equivalent in silicon so far. We show that nanosecond laser single-pass irradiation is capable of producing channel micro-modifications deep into crystalline silicon. With an appropriate shot overlap, a relative change of the refractive index exceeding 10−3 is obtained without apparent nonuniformity at the micrometer scale. Despite the remaining challenge of propagation losses, we show that the created structures form, to the best of our knowledge, the first laserwritten waveguides in the bulk of monolithic silicon samples. This paves the way toward the capability of producing 3D architectures for the rapidly growing field of silicon photonics. © 2016 Optical Society of America OCIS codes: (140.0140) Lasers and laser optics; (140.3070) Infrared and far-infrared lasers; (230.7370) Waveguides; (040.6040) Silicon; (120.5050) Phase measurement; (290.3030) Index measurements. http://dx.doi.org/10.1364/OL.41.004875

The possibility to write waveguides in the bulk of various glasses with femtosecond laser pulses was demonstrated 20 years ago by Davis et al. [1]. The corresponding technique relies on induced structural modifications which are associated with a positive and smooth refractive index change Δn∕n on the order of 10−3 [2]. However, to the best of our knowledge, there is no demonstration of laser inscription of waveguides inside monolithic crystalline silicon (c-Si) so far. Indeed, the natural approach for producing similar structures inside c-Si would be to transpose the femtosecond laser writing technique employed in glasses to its transparency domain (i.e., for wavelengths >1100 nm). Nevertheless, it has been recently demonstrated that the strong plasma defocusing occurring with the use of long-wavelength ultrashort pulses limits the delivered energy density at a level below the permanent modification threshold [3–5]. This represents an important limitation as it completely prevents the development of three-dimensional (3D) femtosecond laser processing technologies in monolithic silicon. A few examples of femtosecond laser-written waveguides in c-Si can be found in the literature, but all have been achieved at 0146-9592/16/214875-04 Journal © 2016 Optical Society of America

an interface with another material such as a-SiO2 [6] or a-C [7]. This is somehow analogous to the lithographic method referred as “silicon on insulator” which is to date the basis technology for all developments in silicon photonics [8]. Despite the excellent performances of these surface technologies, the emergence of a direct laser writing technique inside monolithic c-Si remains highly desirable as it would open a path to the direct fabrication of silicon photonics micro-devices with 3D architectures. Looking at this perspective, it has been recently shown that permanent modifications can be initiated in the bulk of c-Si by nonlinear absorption of nanosecond pulses [9]. However, only uneven modifications exhibiting microcavities have been reported so far [9,10]. These allow addressing applications such as laser dicing, scribing, and marking, [10] but this regime also raises an important question about its level of controllability for envisioning optical functionalizations in c-Si. In this Letter, we report quantitative phase measurements on 1 mm long micro-channel modifications formed deep inside the bulk of monolithic c-Si by repeated illumination with tightly focused infrared (IR) nanosecond pulses. For several tested energies above the modification threshold, we show that one can find an appropriate shot-to-shot overlap so that the modified material gets redistributed in a way it leaves behind an apparently uniform refractive index change Δn∕n exceeding 10−3 . This opens the possibility for direct waveguide fabrication, as confirmed by injecting and detecting guided infrared CW light inside the produced structures. The experimental arrangement employed for producing modifications inside c-Si and in situ mapping of their associate refractive indices is depicted in Fig. 1. The samples are 1 mm thick intrinsic silicon crystals (100 orientation). They are prepared beforehand by cleaving a wafer (from Siltronix) and irradiated through the side surface which, consequently, is optically flat. The nanosecond laser writing setup is schematically represented in Fig. 1(a). The laser source (MWTech, PFL-1550) generates pulses at 1550 nm wavelength, and is operated at a repetition rate of 1 kHz to ensure that there is no heat accumulation on a pulse-to-pulse basis. The pulse duration is τ  5 ns (full width at half-maximum). The laser energy is adjusted using a half-wave plate (λ∕2) and a polarizing beam splitter (PBS). The laser beam is focused by means of a long working distance objective lens [OL (Mitutoyo, Plan Apo,

4876

Vol. 41, No. 21 / November 1 2016 / Optics Letters

Fig. 1. Experimental arrangement divided into three modules: (a) the nanosecond laser writing setup, (b) the IR microscope working in reflection, and (c) the IR quantitative phase microscope. The red, green, and blue colors correspond to 5 ns pulses at 1550 nm, CW illumination at 1200 nm, and 100 fs pulses at 1300 nm, respectively. λ∕2, half-wave plate; PBS, polarizing beam splitter; DM, dichroic mirror; OL, objective lens; LED, light-emitting diode; BS, 50/50 beam splitter; TL, tube lens; PZ, piezoelectric stage; GM, gold mirror; λ∕4, quarter-wave plate.

NIR)] of numerical aperture NA  0.42. Accounting for losses along the optical path, the maximum energy that can be delivered to the target is E max  2.4 μJ. At the focus, the beam is spatially Gaussian-shaped with a theoretical diameter at 1∕e 2 of 2w0  2.3 μm, and a Rayleigh length z R air ≈ 2.8 μm and z R Si ≈ 9.7 μm in air and silicon, respectively. Taking into account the reflections at the sample surface (∼30%), the peak fluence and intensity transmitted in c-Si may reach F max  2E max ∕πw20   80 J∕cm2 and I max  F max ∕τ  16 GW∕cm2 , respectively. The latter is close to the typical values reported for measurable two-photon absorption [11,12]. For precise 3D positioning, the OL and the sample are mounted on motorized stages allowing their displacement along the z axis (i.e., the optical one) and the x; y plane, respectively. In the experiments, channel modifications are longitudinally written in the bulk of silicon with a single pass (i.e., the OL moves upstream toward the incident laser flux) at a speed ranging from 50 nm/s to 2.5 mm/s. The maximum writing depth under the surface is set at 1 mm, ensuring modifications deep inside the bulk of the material without being severely affected by spherical aberrations [13]. The positioning of the beam inside the sample is greatly facilitated by a customized IR reflection microscope [Fig. 1(b)], using a light-emitting diode (LED) at 1200 nm for illumination and an InGaAs array sensor (WiDy SWIR 320U) for image acquisition. For detailed characterizations of the relative refractive index changes Δn∕n associated with the laser-induced modifications inside c-Si (n  3.5 at 1300 nm wavelength), we have specifically developed an IR phase imaging system. Positioned laterally for direct in situ measurements, the setup is depicted in Fig. 1(c) and detailed in [13]. Briefly, it consists of a customized Mach–Zehnder interferometer working at 1300 nm. The sample is placed on the test beam and imaged with a microscopy arrangement built with a NIR objective lens and a second InGaAs array sensor (Raptor, OWL SWIR 640). On the reference arm of the interferometer, the beam is expanded [not shown in simplified Fig. 1(c)], and one mirror is mounted on a piezoelectric stage (Physik Instrumente) for adjustments of

Letter the optical path with a nanometer precision. Accordingly, one can retrieve a quantitative phase image of the modification by acquiring several phase-shifted interference images [14]. Here we apply a simple four-step procedure where the phase is incremented of π∕2 between each acquisition so that the phase is directly given by φ  arctanI 4 − I 2 ∕I 1 − I 3 , where the I k are the four recorded images [15]. The phase image is then numerically unwrapped and flattened so that we can display only the phase shift induced by the modification. To complement this diagnostic, we also systematically acquire amplitude transmission images of the modifications. While the interference images contain also the amplitude information, we prefer to obtain it directly by blocking the reference arm and replacing the coherent illumination by a quartz tungstenhalogen lamp for improved imaging performance. We have first investigated the influence of the writing speed v (and so the shot-to-shot overlap) on the morphology of the structures initiated in the bulk of c-Si at a laser energy of E  2.0 μJ, ensuring that the modification threshold is exceeded [9]. Phase measurements for four typical modifications written at speeds ranging from 0.1 μm/s to 1.0 mm/s are displayed in Fig. 2, and three distinct regimes can be identified. At high speeds [v  1.0 mm∕s, Fig. 2(a)], the morphology of the modification consists of a thin line of positive index change (φ > 0), including dark spots corresponding to a negative index change (φ < 0). These permanent modifications obviously originate from nonlinear absorption of the nanosecond pulses in c-Si leading to temperature and pressure rises in a localized region near the focus [9]. Previous works have shown that the associated hydrodynamic phenomena generally lead to the formation of micro-cavities for similar conditions [16]. Accordingly, with the localized large negative index changes in the image we likely detect these cavities surrounded by a densified material. By repeating shot-to-shot the interactions one creates a pattern that consists of an assembly of these micro-cavities as observed in Fig. 2(a). Here it is informative to note the typical spacing of ∼1 μm between the dark spots (voids) on the image which is comparable to that between each laser shot at v  1.0 mm∕s and for the repetition rate of 1 kHz. An analogous combination of densified regions and microcavities is observed in Figs. 2(c) and 2(d) at low speeds (v  5.0 μm∕s and v  0.1 μm∕s). In these configurations, the spacing between two laser shots is of a few nanometers, and the processes involved in the modifications are more complex as cumulative effects come into play. This results in the formation of micro-cavities that are more randomly distributed. There are also two striking features that are observed for these conditions. First, the modification diameter d clearly increases when the writing speed is decreased (d ≈ 5 μm at v  5.0 μm∕s, and

Fig. 2. Phase images of modifications initiated in the bulk of c-Si for the four indicated writing speeds. The laser energy is E  2.0 μJ.

Letter d ≈ 12 μm at v  0.1 μm∕s). A more extensive study has been carried out in this regime at E  2.0 μJ, showing that the modification diameter scales as d μm  7.4 × vμm∕s−0.2 , for speeds ranging from 50 nm/s to 10 μm/s. The link between d and v reasonably originates from the increase in the amount of material modified by the thermal and structural events with an increasing shot overlap. Second, as highlighted in Figs. 2(c) and 2(d), the lower is v, and the higher is the ratio between the voids and the densified regions. Therefore, voids are more readily produced by the successive laser irradiations of strongly premodified material. Thus, it is relevant to investigate the morphology at intermediate speeds, where one can aim at the most uniform modification of the refractive index. One example of a modification created at a moderate speed (v  0.1 mm∕s) is displayed in Fig. 2(b). Interestingly, the morphology is much more homogeneous than those produced at higher (≥ 1 mm∕s) or lower (≤ 10 μm∕s) speeds. Thus, this shows a relatively narrow process window for refractive index writing applications. According to the high-speed writing results (and single shot experiments not shown here), microcavities must be produced, even for this case [9]. However, one can conclude from the observations that there is an appropriate overlap to get the locally melted material uniformly redistributed. In other words, the motion of c-Si which is melted by each pulse fills the voids left by the preceding ones. We then studied the influence of the laser energy on the produced modifications at high, low, and moderate writing speeds (1.0 mm/s, 10 μm/s, and 0.1 mm/s, respectively). Typical observations at 0.7, 1.2, and 2.4 μJ are depicted in Fig. 3. For each of the three regimes, no change is observed in the general morphology previously described in Fig. 2. In other words, the ratio between the voids and the densified regions is almost the same at the lowest and highest speeds, and no voids are observed at moderate speeds. However, the modification diameter d clearly depends on the energy E. For the specific modifications written at 0.1 mm/s (middle line in Fig. 3) with E ≥ 0.6 μJ, we found a linear behavior between the width and the energy, and no modification observed for E < 0.6 μJ. This value is comparable to the damage threshold associated to silicon melting and reported in [9], determined with similar laser conditions. Hence, by adjusting the laser energy, it is possible to control the width of the modifications produced at 0.1 mm/s without any void formation. This independence of the general morphology of modifications produced at v  0.1 mm∕s from the laser energy has led us to characterize more precisely the phase images obtained at this speed. For instance, for the modification shown in Fig. 2(b), the average phase value is φ ≈ 0.16 rad with a standard deviation of 5.2 × 10−2 rad. The near-uniformity of

Fig. 3. Amplitude images of modifications produced at indicated energies. On each image, the writing speed is 10 μm/s for the top line, 0.1 mm/s for the middle line, and 1 mm/s for the bottom line.

Vol. 41, No. 21 / November 1 2016 / Optics Letters

4877

the modification confers a cylindrical symmetry (along the z axis) that allows us to simply evaluate the average refractive index change by Δn  λφ∕2πd, where λ  1300 nm is the interferometer wavelength and d is the diameter of the apparent modification. By measuring d ≈ 6.2 μm, the corresponding refractive index change is Δn ≈ 5.3 × 10−3 . Other modifications produced in the same conditions exhibit close Δn values, showing the repeatability of the writing procedure. This result is particularly interesting as it is very close to the magnitude of changes reported for the writing of waveguides in glasses with femtosecond lasers [2]. Accordingly, to the best of our knowledge, the written structures represent the first waveguides achieved in the bulk of silicon. Thanks to our lateral imaging setup, we were able to qualitatively confirm this aspect without additional complications. We simply inject CW light emitted at 1550 nm by a laser diode at the entrance of the potential waveguides. Without any other illumination (blocking the microscopy illumination sources) the lateral imaging system creates a scattering image based on submicron nonuniformities which allows us to evaluate how deep IR light propagates inside the structures. An example of the obtained results is displayed in Fig. 4 for a modification written in the bulk of c-Si at a speed of 0.1 mm/s and an energy of 2.1 μJ [Fig. 4(a)]. Thanks to the retrieved value of the index modification, we estimatepthe numerical aperture of the potential waveguide ffiffiffiffiffiffiffiffiffiffiffi NA  2nΔn ≈ 0.19, with n  3.48 at 1550 nm. Then an efficient coupling would require an injection with a focused beam with lower NA. However, because of the modest length (1 mm) of our waveguides obtained by longitudinal writing,

Fig. 4. Observation of light-guiding inside the laser written structures. (a) Amplitude micrograph of a modification produced at E  2.1 μJ and v  0.1 mm∕s. (b) Diffused light inside the same modification while injecting (from right to left) a CW laser diode at 1550 nm on the interface between air and c-Si at z  0. (c) Intensity distribution according to free propagation (nonguided) of the focused CW light. (d) Comparison of the experimental intensity issued from (b) along the modification and the maximum intensity along the dashed line in (c).

4878

Vol. 41, No. 21 / November 1 2016 / Optics Letters

we used an OL of NA  0.45 for injection so that the beam diverges rapidly, and a very small part of the nonguided light illuminates the extremity of the waveguide. Figure 4(b) shows scattered light which is detected along the 510 μm of the displayed structure. While we have seen that phase imaging reveals more details than amplitude imaging, it is striking to note here the nonuniformities detected with this scattering image. However, it is also important to highlight the logarithmic scale used for this image which is possible thanks to the high dynamics of the InGaAs array used in these experiments (>30 dB). Then, to conclude on the detection of guided light through the written structure, one needs to compare the experimental data in Fig. 4(b) to the calculated intensity distribution of the nonguided portion of the light that is tentatively injected inside the structure. As shown by Fig. 4(c), free propagation results in a Gaussian distribution with a waist located at the surface. The intensity distribution (also in logarithmic scale) accounts for the increased Rayleigh length in silicon compared to air (z R Si  nSi z R air). Despite this elongation, these calculations highlight the strong divergence with the chosen NA leading to a decrease in intensity of almost four decades between z  0 and z  −510 μm. For easier comparison, this normalized intensity distribution along the z axis (dashed line) is compared in Fig. 4(d) to the measured intensity in Fig. 4(b) (red circles). At z  −510 μm, the experimentally measured intensity is much higher (∼3 decades) than that expected with free propagation only. Therefore, the scattered light detected at this distance inside the modification in Fig. 4(b) cannot originate from the direct illumination by nonguided light, and we conclude that the modifications guide light. Under the same injection conditions, we have carried out similar experiments on modifications written at high [Fig. 2(a)] and low [Figs. 2(c) and 2(d)] speeds. However, no similar observation was possible because the dark spots with a negative phase induce too significant losses, and rapidly attenuate the injected light. These responses confirm the reliability of our phase measurements to determine the laser writing conditions for optical functionalization of c-Si. While we have concentrated on the refractive index changes that can be induced in c-Si with nanosecond laser pulses, the underlying material science is only partially understood. Subsurface modifications induced by nanosecond laser pulses are usually assumed to result in a melted volume that recrystallizes in more or less disordered regions. For detailed analyzes of the modified silicon crystal structures, one can refer to a recent work of Verburg et al., where transmission electron microscopy (TEM) and Raman spectroscopy were applied [17]. The types of modifications that were found in different regions are: (i) perfectly re-solidified diamond cubic silicon with lattice defects, (ii) amorphous material but also, (iii) microcavities and Si-III/ Si-XII phases evidencing high-pressure conditions. Thanks to phase imaging, we have shown that it is possible to avoid these micro-cavities. The amorphization of silicon is known to lead to an index value significantly higher (n  3.73) than that of c-Si [18]. This is too high for a quantitative agreement with our modifications but it is possible that the positive index variation originates from accumulation of structural defects and disordering of the lattice. Also the strong scattering response seen

Letter in Fig. 4(b) supports the hypothesis of the formation of submicron polycrystalline grains. These were favorable to show the guiding properties by scattering, but will obviously become a major drawback for applications where low–loss waveguides are desired. Finally, the ability to photo-induce positive refractive-index changes in c-Si with focused nanosecond lasers at IR wavelengths has been demonstrated. We will focus our future work on the technological benefits by studying and characterizing in detail the waveguides fabricated by the demonstrated approach (modes, losses, etc.). For simplicity, our demonstration has been achieved with a longitudinal writing geometry, but transverse writing must be implemented to create longer waveguides and for propagation loss measurements. This is also a key aspect to providing the most flexibility for writing inside the materials, as the approach has the potential to open to the 3D the field of silicon photonics that holds promises for the future of ultrafast electronics and quantum computing. Funding. A*MIDEX project (ANR-11-IDEX-0001-346 02); “Investissements d’Avenir” French Government program; Agence Nationale de la Recherche (ANR). REFERENCES 1. K. M. Davis, K. Miura, N. Sugimoto, and K. Hirao, Opt. Lett. 21, 1729 (1996). 2. M. Beresna, M. Gecevičius, and P. G. Kazansky, Adv. Opt. Photon. 6, 293 (2014). 3. V. V. Kononenko, V. V. Konov, and E. M. Dianov, Opt. Lett. 37, 3369 (2012). 4. A. Mouskeftaras, A. V. Rode, R. Clady, M. Sentis, O. Utéza, and D. Grojo, Appl. Phys. Lett. 105, 191103 (2014). 5. E. V. Zavedeev, V. V. Kononenko, and V. I. Konov, Laser Phys. 26, 016101 (2015). 6. A. H. Nejadmalayeri, P. R. Herman, J. Burghoff, M. Will, S. Nolte, and A. Tünnermann, Opt. Lett. 30, 964 (2005). 7. T. H. R. Crawford, J. Yamanaka, G. A. Botton, and H. K. Haugen, J. Appl. Phys. 103, 053104 (2008). 8. M. Lipson, J. Lightwave Technol. 23, 4222 (2005). 9. P. C. Verburg, G. R. B. E. Römer, and A. J. Huis in ’t Veld, Opt. Express 22, 21958 (2014). 10. O. Tokel, A. Turnali, I. Pavlov, S. Tozburun, I. Akca, and F. Ö. Ilday, “Laser-writing in silicon for 3D information processing,” arXiv:1409.2827v1 (2014). 11. A. D. Bristow, N. Rotenberg, and H. M. van Driel, Appl. Phys. Lett. 90, 191104 (2007). 12. D. Grojo, S. Leyder, P. Delaporte, W. Marine, M. Sentis, and O. Utéza, Phys. Rev. B 88, 195135 (2013). 13. Q. Li, M. Chambonneau, M. Chanal, and D. Grojo, “Quantitative phase microscopy of nanosecond laser induced micro-modifications inside silicon,” Appl. Opt. (submitted for publication). 14. J. H. Bruning, D. R. Herriott, J. E. Gallagher, D. P. Rosenfeld, A. D. White, and D. J. Brangaccio, Appl. Opt. 13, 2693 (1974). 15. A. Safrani and I. Abdulhalim, Opt. Lett. 39, 5220 (2014). 16. Y. B. Zel’dovich and Y. P. Raizer, Physics of Shock Waves and High-Temperature Hydrodynamic Phenomena (Academic, 1966). 17. P. C. Verburg, L. A. Smillies, G. R. B. E. Römer, B. Haberl, J. E. Bradby, J. S. Williams, and A. J. Huis in ’t Veld, Appl. Phys. A 120, 683 (2015). 18. M. J. A. de Dood, A. Polman, T. Zijlstra, and E. W. J. M. van der Drift, J. Appl. Phys. 92, 649 (2002).