1. Introduction

Silicon is abundant in the earth’s crust and is widely used in the electronics industry due to its semiconductor properties. It forms a crystalline (c-Si) or an amorphous (α-Si) form in the solid aggregate state that have very different electric, optic and thermal properties [1–3]. Pulsed laser processing of semiconductors has been used for various applications and the resulting surface crystallinity plays an important role for certain applications. Laser cutting and scribing of silicon wafers are widely used in the semiconductor industry [4]. Within the field of optics, laser processing has brought benefits in photovoltaics and photodetectors for efficient light harvesting, achieved by creating parallel and equally spaced laser irradiation lines on silicon substrates that yield an ordered grid of microstructures [5]. Combined laser ablation and modification of silicon have been used to create microlens arrays [6]. Other applications include inducing birefringent properties via periodic nanostructuring [7], creation of nanogratings, where the crystallinity modification is crucial [8], maskless photolithography [9], and formation of precisely defined amorphous regions for applications in the field of optics and photonics [10–12].

Laser pulse duration and energy define the regime of the interaction between light and matter. It has been demonstrated that laser pulses with the pulse energy slightly below the ablation threshold energy cause amorphization of silicon for either nanosecond [13], picosecond [14,15], or femtosecond [16] pulse durations. Laser pulses with the pulse energy greater than the laser ablation threshold cause ablation, leaving behind an ablation crater. The silicon surface layer inside a single ablation crater can either consist of amorphous or crystalline silicon, depending on both the laser pulse duration and energy. Amorphization of the surface layer has been observed using femtosecond pulses with the pulse energy within a certain range and not above it [17,18]. A common effect of pulsed laser processing of silicon in air is the resulting doping with oxygen [19,20] and carbon [20,21] observed at different pulse durations.

We have examined the changes of crystallinity on the whole surface of a microstructure as opposed to the much smaller affected surface due to single pulse absorption. We have created 200x200 µm² sized microstructures using partially overlapping tightly focused picosecond pulses as opposed to larger beam diameter used in other studies [13–21]. Larger surfaces were obtained for a reliable analysis of the cumulative effect of the laser pulse absorption during microstructuring. We have analyzed the silicon surface using micro Raman spectroscopy (MRS), energy dispersive electron spectroscopy (EDS), optical and SEM microscopy. The combined observations have enabled the demonstration of crystallinity and topography control during picosecond laser processing of crystal silicon. EDS results have confirmed low levels of oxidation and carbonization during the laser processing and have verified that the observations with MRS correspond only to different silicon phases.

We have used a custom made picosecond fiber laser source [22] that has enabled the definition of the number of pulses in a burst and the pulse energy. The control of crystallinity and topography is a combined result of the laser parameters and precisely placing multiple bursts of pulses on the material. We have also demonstrated that bursts of pulses increase the ablation efficiency and allow for a precise microstructure depth control. Efficient microstructuring using lower pulse energies when operating with bursts of pulses is an important feature that facilitates the use of fiber lasers that get increasingly complex in design with the need to achieve higher pulse energies. Applications in the field of optics require a precise definition of areas with the chosen shape and crystallinity on the silicon surface. We have demonstrated single-micrometer sized lines and 10 μm high lettering as examples of complex precision-made shapes.

2. Experimental setup

We used an experimental direct laser microstructuring device, based on two key devices: acousto-optic deflectors (AODs) and our own picosecond fiber-amplifier based pulsed laser source [22]. The devices were specially designed for direct microstructuring with nm resolution of the laser beam positioning, using custom driver electronics, software, and optics with AODs (DTSXY400, AA Opto-Electronic) as the positioning system with the maximal repositioning frequency of 100 kHz. The laser source has enabled operation in bursts of 65 ps pulses with stable pulse energies and was externally frequency doubled to a 532 nm wavelength. A burst of pulses is defined as a regime where laser pulses are generated at a high frequency f_i and then bursts of a chosen number of successive pulses are shaped from the pulse train. Bursts are repeated at a significantly lower frequency f_b (f_i = 40 MHz and f_b = 83.3 kHz in our experiment as shown in Fig. 1) and then further amplified to the desired pulse energies.

 

Fig. 1 Schematics of the experiment. On the CAD structure design white areas are to be processed with different laser parameters. The SEM image of the processed sample shows six separate laser processed areas. The scale bar equals to 200 µm.

Download Full Size | PDF

We used a 20x magnification microscope objective (Nikon Plan Fluor 20x/0.5) to focus the laser beam on the silicon surface with the calculated focused beam diameter below 1 μm and the Rayleigh length just below 2 μm. At the beginning of each structuring experiment the silicon surface was located in the laser’s focal plane. Experiments were performed with the repositioning time set to 12 μs corresponding to the repositioning (and consequently also burst repetition) frequency of approximately 83.3 kHz. The laser repetition rate of 40 MHz defined the time delay between consequent pulses in a burst to 25 ns.

A single burst of 1, 5, 10, or 20 laser pulses (PPB – pulses per burst), compensated to achieve equal pulse energies throughout a burst, was delivered on each chosen spot on the material. The average laser pulse fluence on the sample was up to 100 J/cm² at 1 µJ pulse energy and proportionally lower at lower pulse energies. The microstructure was rasterized on a rectangular grid with a variable spacing to ensure a partial overlapping of the absorbed bursts. The spacing was set to 0.4 μm for 1 PPB processing and to 0.8 μm for 5-, 10-, and 20 PPB processing to ensure approximately the same total energy delivered per structure. The spot spacing defines the total number of laser bursts per structure, being approximately 2.5·10⁵ and 6.3·10⁴ for spacing values of 0.4 μm and 0.8 μm respectively. The maximum pulse energy used for multi-PPB bursts was lower compared to the single pulse regime as the average laser power has reached a limit. The positions of the subsequent bursts on the material were chosen randomly as this regime has previously proven to be efficient [23] and to minimize the residual heat from previous bursts. The silicon surface analysis using MRS, SEM imaging and EDS chemical analysis was performed on the bottoms of 200x200 µm² large laser processed areas to minimize the possibility of measurement errors due to the structure edges. MRS was done using a WITec Alpha 300 RS scanning confocal Raman microscope in backscattered geometry using a frequency doubled Nd:YAG excitation laser operating at the 532 nm wavelength. The laser beam was focused through a 100x/0.9 microscope objective and its power at the sample was approximately 7 mW. Each MRS spectrum was obtained on a surface of less than 1 µm² due to the tight focusing of the excitation laser. The signal spectra were acquired in up to 2500 points spread over a larger surface area of each structure to obtain the statistics of different silicon surface types across the processed material. The detection limit of the α-Si layer thickness on c-Si can be determined from the spectrum noise level at 521 cm⁻¹ [24]. Assuming there is no α-Si on the unmodified surface, the detection limit was calculated to be below 1 nm.

Topographic information about the structures was characterized using a Jeol JSM-7600F scanning electron microscope operated with an acceleration voltage of 15 kV. The EDS analysis of the chemical composition of the processed surface was performed with FEI Helios Nanolab 650 also with an acceleration voltage of 15 kV.

3. Results and discussion

Structures were first analyzed in the terms of structure depth and ablation efficiency (graphs in Fig. 2). Ablation efficiency is defined as the amount of ablated material per unit of energy delivered to the material. Bursts of pulses increase the ablation efficiency compared to single pulse ablation at equal pulse energies, which is in line with our previous findings using a tightly focused laser beam [25] and also with findings of other groups using larger laser beams [26–28]. At the same time the total structure depth changed as a function of the total energy delivered to it. Higher ablation efficiency using bursts also significantly increases the structure depth compared to the 1 PPB regime. Larger structure depth leads to the saturation of the ablation efficiency which is visible in all regimes. Due to the tight laser beam focusing the beam spot size increases significantly with depth, leading to lower pulse fluences and thus lower efficiencies. At a structure depth of 25 μm the laser beam size increases by a factor of 10, causing the fluence to drop by a factor of 100, and the ablation efficiency by at least a factor of 4, if the ablation threshold is still reached. The observed changes in the structure bottom topography, as shown in Figs. 2(a)–2(d), have led us to further investigate the final crystallinity after laser processing of pure crystalline silicon.

 

Fig. 2 The highest laser pulse energies resulted in large quantities of ablated material. The total energy is a product of the pulse energy and the total number of pulses and is approximately the same for structures shown above (between 1.03 J and 1.08 J). The number of pulses in a burst varies, resulting in different surface topography. PPB and structure depth were: a) 1 PPB (around 17 μm), b) 5 PPB (around 26 μm), c) 10 PPB (around 25 μm), and d) 20 PPB (around 23 μm). Scale bars are equal to 10 μm. Graphs show the structure depth dependency on the total energy delivered per structure for different PPB regimes and the ablation efficiency as a function of pulse energy. Maximum used pulse energy declines with the increasing number of PPB but the corresponding burst energies get higher.

Download Full Size | PDF

According to the studies published by other groups, we have expected to find amorphous and crystalline regions in the processed surfaces [13–15]. The pure crystalline silicon (c-Si) has a distinct Raman spectrum with a signature one phonon peak at 521 cm⁻¹ and signature multi phonon peaks at 303 cm⁻¹, 432 cm⁻¹, 620 cm⁻¹, 670 cm⁻¹, 820 cm⁻¹, and a broader peak around 960 cm⁻¹ [29] (Fig. 3(a)-A). Amorphous silicon (α-Si) has a clearly different spectrum with two broader peaks around 480 cm⁻¹ and 150 cm⁻¹ [30] (Fig. 3(a)-B). We used our own automatic spectrum-type recognition software based on the recognition of the typical signal peak around 150 cm⁻¹ of α-Si. If there was the peak present, the resulting spectrum was marked as amorphous, and otherwise as crystalline (in Figs. 3(b)–3(e), white areas correspond to spectra of c-Si and black areas to α-Si). Optical attenuation of the 521 cm⁻¹ peak intensity (I₅₂₁(c-Si)) due to an amorphous layer on c-Si (I₅₂₁(α-Si + c-Si)) can be used to calculate the α-Si layer thickness as

 

Fig. 3 MRS results – a) spectra of different types of silicon are shown with typical peaks marked using the symbols shown in brackets: A) c-Si, B) α-Si (■), C) α-Si + c-Si, D) Si-IV + Si-XII (▼), E) strained c-Si (⚫). b-e) Representations of spatial distribution of the amorphous (black) and crystalline (white) silicon surface after laser processing, resulting in b) crystalline surface, c) more than 98% crystalline surface, d) more than 98% amorphous surface, and e) amorphous surface. Each mapping corresponds to a 10x10 µm² silicon surface and consists of the analyzed data from 2500 points. The laser parameters used for each separate mapping are marked on the phase diagrams in Fig. 4. f) EDS mapping, green color represents oxygen, blue carbon, and red silicon. The scale bar equals to 10 µm. g) EDS spectrum, showing only the presence of signature O, C, and Si peaks.

Download Full Size | PDF

Alongside the typical c-Si and α-Si spectra we have observed mechanically strained micro-grains of silicon on the surface (MRS spectrum in Fig. 3(a)-E) and sub-micrometer sized areas of Si-III and Si-XII phases (MRS spectrum in Fig. 3(a)-D). The strained micro grains were recognized by the MRS laser-induced temperature-dependent 521 cm⁻¹ peak shift that is not exhibited by the other Si phases [32]. The peak shifted with an increasing laser power at a rate of 14 cm⁻¹/mW. The Si-III and Si-XII phases are recognized by the signature spectral peaks at 164 cm⁻¹, 353 cm⁻¹, 383 cm⁻¹, 399 cm⁻¹, and 442 cm⁻¹ [33,34]. Most of the anomalies were observed after laser processing with parameters that resulted in a silicon surface with mixed crystallinity. We have cross checked the findings with additional EDS chemical analysis that showed the presence of minor quantities of oxygen and carbon on the processed silicon surface but no other elements (Fig. 3(f) and 3(g), comparable to [17]). No known SiC or SiO compounds have Raman peaks at observed peaks in our spectra [35,36].

Results featuring the crystallinity of the laser processed silicon surface using different pulse energies and PPB regimes are combined in a phase diagram shown in Fig. 4. Laser parameter windows with equal effects are highlighted with uniformly colored backgrounds as following: grey areas represent no detectable effect to the silicon surface using SEM and MRS methods, red areas correspond to the laser induced amorphization with no ablation, green areas correspond to a silicon surface with mixed crystallinity, pink areas correspond to a crystalline silicon surface and blue areas correspond to simultaneous amorphization and ablation. Near the borders of respective regions, the surface is not necessarily 100% crystalline or amorphous. The areas are marked in such a way that the data points inside corresponds to at least 98% of surface coverage by the expected type. For example, at 10 PPB and pulse energy 0.7 µJ the laser processed silicon surface is at least 98% crystalline. The data points measured are represented with different symbols and are marked according to the prevalent silicon crystallinity type also in the cases with the present anomalies mentioned before.

 

Fig. 4 Phase diagrams of silicon crystallinity after pulsed laser processing, plotted in the parameter spaces of the number of pulses in a burst (PPB) versus pulse energy and total energy per structure. The data points corresponding to the representations of the silicon surface type spatial distributions shown in Figs. 3(b)–(e) are marked with the matching labels.

Download Full Size | PDF

The phase diagram shows roughly similar transitions through the observed regimes regardless of the number of PPB chosen, starting at no effect at low energies through amorphization and onward to ablation. However, the use of bursts in combination with tight laser beam focusing enables an unprecedented option of controlling the resulting silicon surface crystallinity. The use of tight focusing in combination with 65 ps pulses leads to an averaged effect over the whole surface due to comparable typical distances of electron and heat transport to the laser spot size as opposed to the creation of amorphous rings with larger laser beam diameters [12,16,37]. The amorphization observed at a range of sub-threshold pulse energies is in agreement with studies of other groups [9,13]. Use of single pulse bursts at energies just above the ablation threshold has resulted in structures with amorphous surface. A single pulse still induces a rapid-enough quench even at energies above the ablation threshold that results in reliable amorphization [17], while the longer total duration of a burst of 5 or more pulses does not induce the rapid quench anymore. At high pulse energies the opposite effect emerged as multi pulse bursts resulted in structures with a crystalline surface, but no such regime was observed within the experimental parameter range using single pulse bursts. Extrapolation suggests that such regime might exist at pulse energies around 5 µJ. Energetic bursts of pulses result in higher average fluences delivered to a single spot on material. We propose this as the leading cause to a slower solidification dynamic resulting in a fully crystalline surface, which is in agreement with [17]. In between, a large laser parameter window remains where material was ablated with the resulting surface a mixture of different silicon phases. This window opens a chance to work at the best possible setting for either achieving the desired depth of structure, bottom topography, or the available laser output if the surface crystallinity control is not needed. The blue and pink areas represent the significant laser parameters for precise surface crystallinity control, enabling also the fine tuning of structure topography as demonstrated in Figs. 2(a)–(d). We have observed that surface roughness increases with an increasing number of pulses in a burst during laser ablation processing at equal total energy used.

The choice of correct parameters, as indicated on the phase diagrams in Fig. 4, defines the light-matter interaction regime. We have used 10 PPB mode and pulse energy around 0.025 µJ to create the conditions that cause selective amorphization of the crystalline silicon surface. Pulse overlapping was set to 0.2 µm for line objects and 0.4 µm for lettering which yields approximately the same total fluence. The results are in form of approximately 1 µm wide lines and 10 µm high lettering with edges smoother than 200 nm measured as the peak to peak variation from an ideal shape. The structures are visible on the optical images (Fig. 5) due to reflectivity difference, as bulk c-Si reflects approximately 37% of visible light while bulk α-Si reflects approximately 41% [1,31]. The reflectivity of α-Si layer on c-Si substrate is higher compared to pure bulk c-Si for the typical α-Si layer thickness range obtained in our experiment [12,14]. The demonstrated process capabilities present a valuable tool to selectively change the optical properties of the material on the given scale.

 

Fig. 5 Optical images of precisely defined amorphous regions in the form of lines (horizontal, vertical, spiral) and lettering, on an otherwise crystalline silicon surface. Scale bars are equal to 5 µm. The graph shows reflected light intensity values across the horizontal α-Si lines averaged along the whole line length.

Download Full Size | PDF

4. Conclusion

We have demonstrated and analyzed a novel approach to the microstructuring of bulk silicon with a tightly focused laser beam. The combination of tight focusing and precise spot position control achieved by the use of AODs have led to uniform processing and homogeneous surface properties over either larger silicon surface areas or precisely defined single-micrometer sized structures. Variation of pulse energy and the number of picosecond pulses in a burst have enabled full control of structure surface crystallinity.

Fast and efficient micro ablation in combination with precise control of the surface crystallinity is an important step to the integration of direct laser processing in more complex manufacturing processes. Our research has opened a possibility to fully exploit reliable and stable fiber laser sources with rather low pulse energy output due to the increased efficiency of ablation, increased average powers due to the use of bursts and the achieved final structural properties of silicon only exhibited in the multi pulse per burst regime.