Introduction

Focused ion beam (FIB) technology has become increasingly attractive since the introduction of the liquid metal ion source (LMIS) [1,2]. Today, FIB is an appealing tool in different research areas: in the semiconductor and electronics industry [3], life sciences [4,5], optics and photonics applications [6–14] and micro-electromechanical systems (MEMS) [15,16].

The diffusion of FIB technology for different applications has largely been due to its several features, namely choice of ion species [17], variable ion current [8,9,15] and ion energy and injected gases [3,15]. The same tool allows imaging [4,5], (gas assisted) milling [3–5,8–19], gas assisted deposition [3,15] and implantation [7] of a selected sample. Moreover, a wide range of materials such as semiconductors [8,10,17,19], metals [3,6,9], soft materials [4,5] and insulators [7,11,12,18] can be processed by FIB. In addition, it offers a maskless fabrication process combined with the possibility of very fine focusing and high resolution. The combination of all these factors allows the fabrication of structures with complex geometries [10,18], small features down to sub-5 nm dimensions [5,19] and high aspect ratios [6]. However, when energetic ions hit the target, a variety of ion-solid interactions can occur, causing sputtering, surface amorphization, implantation, swelling, deposition, backscattering and nuclear reactions. In addition, FIB milling of III-V compounds results in appreciable residual group-III species on the surface, often in the form of droplets [17,20]. This phenomenon is well known for Ga⁺ ion beam milling, which is the most commonly used technique [21]. Attempts to avoid the formation of the III-material-rich droplets have been made by milling at cryogenic temperatures but without success [22]. After the FIB milling it is possible to partially recover the samples by thermal annealing, usually performed at temperatures ranging from 200 °C to 600 °C [13].

Among different III-V compounds, gallium phosphide (GaP) has a high second order nonlinear coefficient [23], high thermal conductivity [24] and a broad transparency range [25], which make it an excellent material for nonlinear optical applications. Moreover, it has a high refractive index [25], which enhances the rapid variation of electric field across GaP/ air interfaces, contributing to the strong surface optical nonlinearity. GaP has been widely employed for photonics applications, especially in its nanopillar and nanotube shapes [26–30], but other examples of geometries are reported as well [10, 31–33].

In this paper, we present the fabrication of GaP multilayer slab waveguides (MSWs) by FIB. A post-fabrication step in which the FIB milled MSWs are treated by pulsed laser irradiation in ambient air is used to recover the optical properties by oxidising the surfaces gallium droplets. By measuring the second harmonic generation (SHG) of light from the MSWs, we demonstrate that this fabrication procedure guarantees high second-order nonlinear optical properties of the structures, including but not limited to the surface properties.

Focused ion beam fabrication

A dual-beam FIB machine FEI Nova 200 has been used for fabricating the structures. This system contained both a tilted ion beam column and a vertical electron column, which allowed real time monitoring and non-destructive characterization by scanning electron microscope (SEM). The ion beam was generated from a Ga⁺ liquid metal source. For the experiments reported here, the acceleration voltage was set at 30 kV; the FIB milling currents at 10 pA and 50 pA corresponding to a nominal beam diameter of 12 nm and 19 nm, respectively. We note that the FIB current can affect the implantation characteristics as well as the amount of gallium generated. However, our primary purpose to use the higher current was to fabricate structures in a shorter time and/or with larger heights. The dwell time was defined at a value of 20 μs. Samples of undoped (100) GaP were used.

Two types of structures have been fabricated (Fig. 1). The first type (I) (Fig. 1(a)-1(b)) was a MSW fabricated at ion beam current of 10 pA, consisting of parallel lines of width w_n = (0.45 ± 0.05) μm and length l = (4.90 ± 0.05) μm, 50% duty cycle and height h = (1.10 ± 0.01) μm. Here, w_n represents the average between the top and at the bottom widths, as measured from the top-view SEM images. MSW (I) was made of parallel slabs with the same dimensions and spacing, as a standard waveguide. The interest for the structure derived from the possibility to compare the optical response in the same conditions for each of the slabs. The second MSW (II) (Fig. 1- (c-d)), has been realised in two versions by using ion beam currents of 10 pA and 50 pA. They were fabricated as a progression of widths of the slabs to investigate the dependence on geometrical dimensions on surface and bulk optical nonlinearities. The structure fabricated with ion beam current of 10 pA ranged from w_n = (0.04 ± 0.01) μm to w_n = (0.40 ± 0.05) μm, length l = (5.10 ± 0.05) μm and height h = (0.30 ± 0.01) μm. On the other hand, MSW (II) fabricated at 50 pA, ranged from w_n = (0.03 ± 0.01) μm to w_n = (0.42 ± 0.07) μm, length l = (5.05 ± 0.05) μm and height increasing from h = (0.60 ± 0.01) μm to h = (1.30 ± 0.01) μm corresponding to the smallest and the largest slab, respectively. The best aspect ratio achieved was approximately 20. Both types of structure were patterned parallel to the crystallographic axis of GaP wafer, i.e. 45° rotated respect to the cleaving edges of the samples.

 

Fig. 1 Fabricated GaP multilayer slab waveguides. Indicated scale bars 1 µm. (a) SEM picture of the structure type I (fabricated at 10 pA FIB current). The insets show magnified view of the structure where w_n represents the average width of the waveguide. (b) Schematics of type I. (c) SEM picture of the structure type II (fabricated at 10 pA FIB current). (d) Schematics of type II.

Download Full Size | PDF

Analysis of second-order optical nonlinearity

For GaP and all the crystals with $\overline{4}3$m symmetry, the only nonzero components of the second-order nonlinear susceptibility tensor are d₁₄ = d₂₅ = d₃₆ = d. Therefore, the second-order nonlinear polarization along the crystallographic axes are as follows [30]:

Furthermore, at the surface, the inversion symmetry is broken, resulting in new nonzero elements of the nonlinear susceptibility tensor. Both material and field discontinuities contribute to the surface nonlinearity. If the x-axis is orthogonal to the surface of the crystal with $\overline{4}3$m symmetry, the second-order nonlinear susceptibility tensor [29] can be defined as:

In case (a) φ_PUMP = 0°, tensor in Eq. (4) leads to the second-order nonlinear polarization at the surface P_x⁽²⁾ = d^S₁₁ E_x².

Considering the symmetry of the transverse and longitudinal components for the guided pump in the mode TM₀ and case (a), the nonlinear polarization density can excite two modes at the SHG wavelength, TM₁ and TE₁, respectively related to the surface and the bulk contributions to the nonlinear polarization. Since P_y⁽²⁾, results in a SHG light with polarization perpendicular to the pump and P_x⁽²⁾, results in a SHG light with the same polarization of the pump, it is possible to identify bulk and surface contributions by polarization measurements.

In case (b), the pump is linearly polarized at 45° respect to the crystallographic axis. Therefore, it is guided in a superposition of TM₀ and TE₀ modes. Hence, the polarization of the SHG light depends on the phase-matching condition between the pump guided in TM₀ and TE₀ modes and SHG light guided in TM₁ and TE₁.

Second-harmonic generation measurements

A Ti:Sapphire laser with 100 fs pulses and a central wavelength of 840 nm at 82 MHz repetition rate was used as the pump for the second harmonic generation in the transmission configuration setup reported in Fig. 2. The laser beam passed through a rotatable half-wave plate that allowed to rotate the linear polarization of the pump and, successively, was focused by a 50x Olympus microscope objective, with NA = 0.5. The incident pump propagated along the crystallographic z-axis of the sample with an average diameter of 2 μm on the top surface. The generated SH light was collected by a 50x Olympus objective, with NA = 0.5 and the collimated beam was passing through a Schott Glass band pass filter BG39 (blocking pump wavelength) and a rotatable polarizer. This setup allowed the measurement of SHG light in two different configurations: (i) parallel to the pump and (ii) orthogonal to the pump, which correspond to φ_SHG = φ_PUMP and φ_SHG = φ_PUMP + 90°, respectively. The SHG light was coupled to a multimode fiber and guided to a monochromator, set at 420 nm. Finally, a Perkin Elmer silicon avalanche photodiode was used to measure the intensity of the SHG light.

 

Fig. 2 Optical setup for the polarization measurements of the SHG light.

Download Full Size | PDF

As a post-processing step, the same pump laser was used to expose the samples with an average power of 150 mW and a constant polarization (case (a), φ_PUMP = 0°).

For short time exposure, below 5 min, MSW (I) was tested. In this case, the SHG signal increased above 200% from the initial value (Fig. 3-a). Since the MSW(I) was 450 nm wide, no contribution from the surface to the SHG light was observed.

 

Fig. 3 SHG intensity for pulsed-laser exposure with average power P_av = 150mW. (Blue line) bulk contribution to the SHG light, (red line) surface contribution to the SHG light. (a) MSW (I) (fabricated at 10 pA FIB current), waveguide width w_n = (450 ± 50) nm. (b) MSW (II) (fabricated at 50pA FIB current), waveguide width w_n = (280 ± 10) nm.

Download Full Size | PDF

In contrast to MSW (I), for MSW (II) fabricated at 50 pA it was not possible to measure any SH generated signal. To recover the structure, additional process steps were necessary. The process cycle consisted of fast scanning of the entire structure with the laser (~2 s per spot, 150 mW), oxygen plasma cleaning and hydrofluoric acid (HF) wet etching. After two such process cycles SHG measurements were performed and Fig. 3(b) shows SHG light intensity as a function of laser exposure time (150 mW average power). In this case, the polarizer for SHG light was rotated from parallel to orthogonal configurations in each measurement. This allowed the observation of contemporary evolution of the bulk and the surface contributions to the SHG light. The graph presented in Fig. 3(b) shows increase of the bulk contribution above 200% of the initial value. At the same time, the surface contribution to the SHG light stabilized at about 25% of the initial value. It was possible to observe an increasing bulk contribution to SHG light and a reduction of the surface contribution, which we attributed to an oxidization process. Oxidation of the surface leads to reduced electric field gradient and modification of material discontinuity at the interface between gallium phosphide and air.

Concerning the SHG measured in Fig. 3, we note that in Fig. 3(a), the intensity fluctuations are due to dynamic thermal effects, such as absorption by gallium droplets, chemical modifications and possible redistribution of surface gallium. In the case of Fig. 3(b), the reduction of the intensity after ~37 min is due to laser-induced damage of the waveguides.

Nonlinear properties of the fabricated samples, specifically the bulk and surface nonlinear properties, where tested by polarization measurements. The polarization plots allowed to identify the bulk and surface contribution to the SHG light. In Fig. 4, the bulk contribution to the SHG light is identifiable in the orthogonal configuration (blue), whereas the surface contribution is identifiable in the parallel configuration (red). In Fig. 4(a), the orientation of the crystallographic axes with respect to the measurements is presented. All the polarization measurements were recorded for the average laser pump power of 50mW.

 

Fig. 4 Polarization plots of SHG light (all in the same scale). In red, parallel configuration, in blue, orthogonal configuration. Laser average power P_av = 50mW. The angle represented in the axis is φ_SHG. (a) Orientation of the crystallographic axes with respect to the polar plots. (b) Substrate. (c-k) MSW (II) (fabricated at 10 pA FIB current).

Download Full Size | PDF

Measurements of the SHG light from the substrate indicate the orientation of the crystallographic axis (Fig. 4(b)) and reveal the four-fold symmetry of the crystal. In fact, in the orthogonal configuration, the contributions P_x⁽²⁾ (Eq. (1)) for φ_SHG = 0°, 180° and P_y⁽²⁾ (Eq. (2)) for φ_SHG = 90°, 270° are clearly visible. In the parallel configuration, instead, the contribution P_z⁽²⁾ (Eq. (3) is dominant.

In Fig. 4(c)-4(i) the dependence of the intensity and contributions to the SHG light from the different widths of the slab waveguides, in structure (II) (10pA) is shown. Measurements were done after the post-processing step consisting of the laser exposure for less than 5 min. Due to the laser spot diameter (2 μm) the contributions to SHG light are presented for the average waveguides width (w_av). The maximum error of the waveguide width (Fig. 4(c)-4(i)) is defined by the smallest and the largest slab in the area covered by the laser spot during the measurement. For w_av = 250 nm (Fig. 4(f)), the transverse contribution of the bulk, given by P_y⁽²⁾ (Eq. (2)) for φ_SHG = 90°, 270°, is dominant. This is due to a higher value of the field E_z, which indicates the strongest confinement of the electric field. Hence, for w_av = 210 nm (Fig. 4(g)), the surface contribution to the SHG light is dominant. This is visible at angles φ_SHG = 0°, 180°, indicating contribution from nonlinear polarization component P_x⁽²⁾. Furthermore, for w_av = 150nm (Fig. 4(h)), it is possible to distinguish the dominant longitudinal component (P_z⁽²⁾). When φ_PUMP = 45°, both E_x and E_y contribute to P_z⁽²⁾ (Eq. (3)). Due to the waveguide geometry, in the far field, P_z⁽²⁾ is responsible for the SHG polarization normal to the side surfaces of the waveguide. This results in the four lobes being twisted by ± 10° with respect to the substrate (Fig. 4(b)).

In order to investigate the influence of longer laser exposure time (presented in Fig. 3(b)) on waveguide properties, polarization measurements on MSW (II) (50 pA, w_av = 280 nm) were performed (Fig. 5) at laser average power of 50 mW. Polarization plots before (a) and after (b) 45 min exposure show at φ_SHG = 0°, 180° the surface contribution to the SHG light (in red) reduced to about 20% of the initial value, where at φ_SHG = 90°, 270° the bulk contribution to the SHG light (in blue) increased to above 200%. At the same time, the change of the overall shape of the bulk contribution is visible. It is a sign of the modification in the confinement of the light inside the structure. This is attributed to the reduction of the effective waveguide width due to laser-induced local heating effects, resulting in surface oxidation and material redistribution. In addition, the increase of the bulk contribution indicates lower absorption, where decreasing surface contribution is due to the surface oxidization. Note, however, for larger currents (as for 50 pA) the sidewalls of the waveguide are less steep than for the same structure patterned at lower current (10 pA). This results in a smaller waveguide width at the top. Therefore, the plots presented in Fig. 5 have similar shape to the one in Fig. 4(h)-4(i)) corresponding to smaller waveguides, w_av = 150 nm and w_av = 125 nm.

 

Fig. 5 Polarization plots of SHG light (scale reduced 9 times with respect to Fig. 4) in MSW (II) (fabricated at 50 pA FIB current), w_av = (280 ± 10) nm. In red, parallel configuration, in blue, orthogonal configuration. Laser average power P_av = 50mW. The angle represented in the axis is φ_SHG. (a) before and (b) after the 45 min laser exposure presented in Fig. 3 (b).

Download Full Size | PDF

Compositional analysis

After the polarization measurements, the samples were investigated with SEM (10 kV, Ultra 55 Gemini, Zeiss). The imaging of the sample revealed the presence of newly formed superficial layer made of nanoparticles (<100 nm), at the same places where the GaP MSW were exposed to the pulsed-laser beam (Fig. 6(a)). In addition, energy dispersive X-Ray (EDS) compositional analysis (10 kV, 2k channels, 60 s, Ultra 55 Gemini, Zeiss) showed a higher presence of oxygen there (Fig. 6(b)).

 

Fig. 6 Compositional analysis of MSW (I). (a) SEM image of the structure. Position A (red) and B (blue) correspond to areas with and without laser exposure, respectively. Scale bar 1µm. (b) Energy dispersive X-Ray (EDS) graph.

Download Full Size | PDF

To remove the negligible residue of carbon (C) present on the samples (Fig. 6(b)), a cleaning step with oxygen plasma (10 min, 500 sccm of O₂, 1000 W, TePla Model 300 Plasma System) was made. This step did not affect in a substantial way the oxidization of the structures; no oxide layer was found in the areas not exposed to the laser.

Since laser deposition technique has been extensively used for growing oxides, such as gallium oxide (Ga₂O₃) [34–37] and titanium oxide (TiO₂) [35], we attributed the detection of oxygen as the result of gallium oxide formation on the surface. In order to crosscheck the presence of gallium oxide, the samples were immersed in 50% hydrofluoric acid (HF). After 1 min, this process effectively removed the oxide (with the reaction Ga₂O₃ + 6HF→2GaF₃ + 3H₂O [34]), leaving small “cleaning” marks on the surface (Fig. 7(b)). Therefore, laser exposure is the essential step, since the following HF wet etching, is not effective otherwise.

 

Fig. 7 SEM image of MSW (II) fabricated at 50pA FIB current. Scale bar 1μm. (a) After the fabrication with FIB. (b) After the laser, the oxygen plasma and HF etching.

Download Full Size | PDF

For the structures made with 10 pA, the localized exposure to the laser beam was less than 5 min (as shown in Fig. 3(a)) in order to have a detectable SHG light.

On the other hand, for samples fabricated with higher currents (50 pA) it was not possible to measure any SHG light even after laser exposure times longer time than 5 min. However, in this first exposure, the laser beam stimulated the oxidization of the surface, which was then etched by HF. The same sample was processed similarly once more. Only after that procedure, it was possible to measure the SHG light from the structure. In Fig. 7, it is clearly visible the difference betweenthe surface layers before (a) and after (b) the complete process, consisting of laser exposure, oxygen plasma and HF wet etching. The necessity to perform extra steps in the case of the structure fabricated with higher current can be explained by the higher density and the larger dimensions of the gallium droplets and the higher redeposition rate of the milled material.

Some side effects of using the presented post-process steps were also observed. First, the formation of the gallium oxide continued after the laser exposure. Imaging the samples with the SEM (10 kV, Ultra 55 Gemini, Zeiss) several weeks after the measurements, it was clearly visible that an increased portion of the surface was covered by the gallium oxide. The oxidization was observed also in the areas shortly exposed to the laser, e.g. along the paths from one measurement point to the next. In those cases, the oxidization was not visible just after the measurement. However, the process can be stopped by HF wet etching, since the oxide layer (corresponding to defects inside the GaP structure) is removed. On the other hand, pumping the fabricated structures with higher average laser power (>150mW) or for longer time irremediably damaged the structures.

Conclusions

Multilayer slab waveguides were fabricated on a single-crystal (100) gallium phosphide wafer using focused ion beam milling at 10 pA and 50 pA currents and 30 kV acceleration voltage. This process gave good quality structures (smallest features 30 nm with aspect ratio 20:1) but with superficial droplets of gallium, which limited the optical properties of the waveguides. To solve this problem, a pulsed-laser beam at 150mW was used to expose the nanostructures surface and convert the gallium-rich surface into gallium oxide. In order to test the quality of the structures for potential applications in nanophotonics, we analysed the second-harmonic generation response. For the structure patterned at 10 pA and laser exposure of less than 5 min, we could observe bulk and surface contributions to the optical nonlinearity. Interestingly, for the structure patterned at 50 pA, the repetition of the process consisting of pulsed-laser exposure, oxygen plasma and wet etching with hydrofluoric acid, was required to improve the quality of the waveguides. Moreover, while a longer laser exposure duration reduced the optical losses further, it also increased surface oxidation thereby decreasing the surface optical nonlinearity. Furthermore, the presented method can be used as a post-process technique to recover the nanostructures fabricated in III-V materials with focused ion beam. The reported method and the possibility to monitor the oxidization on the surface can open doors to a variety of different applications which require high aspect ratio structures in III-V materials or gallium oxide nanoparticles on complex structures.