1. Introduction

Second-order nonlinear (NL) optical properties are forbidden in materials which present inversion symmetry on a macroscopic scale, such as glasses. However, it is possible to induce a second-order nonlinear susceptibility in bulk glass using specific treatment like optical assisted poling or thermal poling [1–3]. The latest method has proved to be an efficient technique for producing reproducible χ⁽²⁾ susceptibilities in oxide glasses. It consists in heating the glass sample to a suitable temperature (below the glass transition temperature) and in subsequently applying a high electric field across it. After some minutes, the glass is cooled down to room temperature before removing the electric field. On a microscopic scale, two mechanisms are mainly proposed to explain the origin of the nonlinearity occurrence in oxide glasses. The first mechanism is related to migration of ionic species under the applied electric field, which leads to the creation of a permanent electric field E_DC. The coupling of E_DC and the third-order susceptibility (χ ⁽³⁾) of the glass gives rise to an effective ${\chi }_{\text{eff}}^{\left(2\right)}$ through equation ${\chi }_{\text{eff}}^{\left(2\right)}$∝χ ⁽³⁾.E_DC. The second mechanism is related to reorientations of polar bonds or hyperpolarizable entities during the poling treatment resulting in induced macroscopic second-order nonlinearity. It is well established that the χ ⁽³⁾ of chalcogenide glasses is about two orders of magnitude larger than that of silica. Thus large SHG efficiencies can be expected in such glasses [4–7]. Indeed, the creation of second-order nonlinearities in chalcogenide glasses is of great interest due to their interesting properties: large infrared spectral range of transparency (up to 10 or 20 µm depending on the composition), low phonon energy, photosensitivity, high linear and nonlinear refractive indices and the possibility to perform optical waveguides. Therefore, these materials present great potential for the conception of electrooptic modulators in the infrared (IR) spectral region.

The purpose of this study is to investigate the creation of second-order nonlinearities in sulfide bulk glasses. In this paper, we report the effect of a thermal poling in a glass sample with nominal composition Ga₅Ge₂₀Sb₁₀S₆₅. A large SHG effect was obtained with a poling process in controlled atmosphere.

2. Experimental conditions

2.1. Glass synthesis and characterization

High purity elements were used for glass preparation, germanium, gallium, antimony and sulfur 5N. Despite of the purity of the commercial material, sulfur can be polluted by water and carbon. Water can be eliminated by heating sulfur under dynamic vacuum at 125°C and carbon by sulfur distillation at 350°C. After purification the raw materials were placed in a silica tube, which is sealed under vacuum (10^-5 mbar). After sealing, the batch was slowly heated to 800°C and homogenized in a rocking furnace at this temperature during 12 h. A glass rod was obtained by cooling the silica tube in air. It was then annealed near the glass transition temperature for 30 min before slowly cooling down to room temperature. Several glass plates about 1 mm in thickness and 20 mm in diameter were obtained from a same batch. For optical characterizations, the samples were optically polished to get sides as plan and parallel as possible.

Thermal analyses were carried out on single glass chips, about 50 mg, in sealed aluminum pans in the temperature range of 25–500°C. The measurements were performed at a heating rate of 10°C.min^-1 by means of a differential scanning calorimeter with an accuracy of ± 2°C. Analysis using Energy Dispersive X-Ray Spectroscopy (EDS) allowed the identification of the elemental composition of the studied samples imaged in a Scanning Electron Microscope (SEM) and permitted to control the stoechiometry of the obtained glasses.

2.2 Poling treatment

Before poling treatment, a thermal gravimetric analysis (TGA) of a glass chip at 300°C during 30 min has been performed in air and N₂ atmospheres. A slight loss of mass due to the surface oxidation in the case of air experiments has been shown. Consequently, poling treatments were realized under N₂ gas to avoid this phenomenon. Thermal poling was carried out with poling temperatures of 230, 250, 270 and 290°C. Thermal equilibrium duration before applying the voltage was fixed to 1h30. The applied voltage was 4 kV. The electric field was applied across the sample by means of two silicon plate electrodes. An air gap (about 125 µm in thickness) has been added between the sample and the electrodes using silica fibers, in order to limit the current flow in the glass. Once the field has been applied during 30 min, the sample was cooled down to room temperature and the high voltage was removed.

2.3 Optical characterization

Visible-IR spectra were recorded at room temperature between 500 and 3200 nm using a CARY spectrophotometer (Varian). Linear refractive indices were measured using an experimental technique based on the measurement of the sample transmission for different lengths with the help of an IR spectrophotometer (Lambda 19 UV-Vis/NIR of Perkin-Elmer). Readers can consult Ref. 8 for further details about this method. For nonlinear measurements, all samples were characterized using a classical Maker Fringe (MF) experiment with a pump beam at a wavelength of λ_ω=1.904 µm [9]. This wavelength is chosen to avoid the absorption of the SH wave (λ_2ω=1.904/2 µm=952 nm) by the glass, whose bandgap wavelength is about 583 nm. This experiment consists in recording the second harmonic (SH) power as a function of the incident angle of the pump beam. The pump beam is produced by an Optical Parametric Oscillator (type I, BBO crystal in a resonator), itself pumped by a Nd-YAG pulsed laser operating at 532 nm with a 10 Hz repetition rate. As the BBO crystal produced two beams, the signal at 738 nm and the idler at 1.904 µm, appropriate filters have been used to separate the idler from the signal so that only the idler beam reaches the sample. The refractive indices are estimated to be n_ω=2.244 and n_2ω=2.285, the associated coherence length L_C being about 11.5 µm. A α-quartz crystal is used to calibrate the experiment and to measure the χ ⁽²⁾ susceptibility.

3. Results

The glass composition has been analyzed after elaboration and proves to be homogeneous from sample to sample (Ga (4,5±0.5) at.%, Ge (21,5±0.5) at.%, Sb (10,5±0.5) at.%, S (63,5±0.5) at.%). The discrepancies between the theoretical composition and the real one are almost identical for several glasses.

 

Fig. 1. MF pattern obtained for two samples of Ga₅Ge₂₀Sb₁₀S₆₅ from two different batches. The samples have been poled under the same conditions.

Download Full Size | PDF

The reproducibility of elaboration process was also tested using samples from two different batches through SHG measurements. Fig. 1 shows the SH signals recorded after poling at 230°C on samples from the two batches. As it is shown, the differences between the Maker fringes patterns are negligible. As a consequence, poling treatment gives reproducible results for two samples from two different batches. Different poling temperature conditions were studied. Fig. 2 shows the recorded SH signals obtained after poling at 230° C, 250° C, 270° C and 290° C. The SH signal increases when the absolute value of the pump beam incident angle increases and reaches a maximum around 60°. It decreases then to zero due to Fresnel losses at high angle. The more significant SH signal is obtained for a temperature of 270°C. However, SEM analysis of the cathode and anode sides of glass samples indicates the presence of surface damage for samples poled at temperature higher than 230°C. To avoid this phenomenon, optical characterizations have been performed on samples poled at a temperature of 230°C.

 

Fig. 2. SH signals obtained after poling at 230°C (full squares), 250°C (open squares), 270°C (full triangles) and 290°C (open triangles). The applied voltage is 4 kV and the poling duration is 30 minutes.

Download Full Size | PDF

The characterization of the NL spatial distribution was performed using a method described in Ref. 10. Basically, the method consists in recording the SH signal as a function of the depth under the anodic side of the poled sample. To this end, thin layers of glass have been removed using a surface etching method involving sodium hydroxide (NaOH) solution. The anode side poled region was partially immersed in NaOH for controlled etching periods. After every etching stage, the protected anodic poled region allowed a control of the initial SH signal, and the remaining SH power was measured for a fixed incidence angle (60°) of the pump beam. The removed thickness for each step was measured by a profilometer technique (DekTak 3 ST surface optical profilometer, with a vertical resolution of about 10 nm). Fig. 3 presents the SH signal as a function of the removed thickness (black dots) and a curve fit of the experimental data (black line). The best fit was obtained using the following function:

where x is the depth across the sample, a, b and c are constants which are respectively equal to 225.7, 10.87 and 2.02.

 

Fig. 3. Remaining SH signal after successive etching operations. The filled dots represent experimental points. The black line represents the best theoretical fit.

Download Full Size | PDF

As displayed in Fig. 3, the SH signal decreases monotonously as a function of the removed layer thickness of glass and becomes negligible after having removed approximately 15 µm. It is noticeable that the SH signal decreases slowly in the first 3 µm. This behavior implies that the NL layer is buried inside the sample.

4. Discussion

Several mechanisms can be considered to explain the establishment of a nonlinearity in thermally poled Ga₅Ge₂₀Sb₁₀S₆₅ glass. In silica glasses, the migration of mobile cations, like Na+, towards the cathodic surface to the detriment of the anodic one is known to be at the origin of the SHG [11]. As a consequence, a depletion layer (negatively charged) is induced within a few micrometers near the anode. This accumulation of negative charges leads to the creation of a high electric field E_DC near the anodic side of the sample [11]. The result presented in Fig. 3, i.e. the fact that the SH is generated only within the first 15 µm, supports the hypothesis of the electric field E_DC creation. This mechanism could be followed by the reorientation of the lone pairs of the sulfur and antimony atoms under the induced electric field E_DC, which can also induce nonlinearities via the microscopic hyperpolarizability of such entities.

The determination of the nonlinear spatial profile is obtained using the “layer peeling” method described in Ref. 10. This method allows to compute the spatial NL distribution χ ⁽²⁾(x) using the experimental SH power evolution as a function of the depth. Contrary to ref. 10 which describes a direct data reconstruction, the fit (1) of the experimental data has been used to retrieve the induced nonlinearity. It leads to Fig. 4 which displays the χ ⁽²⁾ spatial distribution achieved in a poled chalcogenide sample. As can be seen, the NL profile is almost Gaussian. Its maximum is buried about 9.4 µm inside the sample and its full width at half maximum is estimated to be about 6.5 µm in thickness. The fact that the NL layer is buried inside the sample implies that partial neutralization mechanisms could take place during or after poling: this may corresponds to the emission of negative charges from the maximum field zone corresponding to current spikes detected in oxide poled glasses (electron or sulfur in our case) and/or to the injection of positive mobile ions (H₃O⁺ or N⁺).

 

Fig. 4. Reconstructed nonlinear χ ⁽²⁾ susceptibility as a function of the depth under the anode.

Download Full Size | PDF

The absolute measurement of the χ ⁽²⁾ susceptibility is then deduced from the experimental MF pattern. In first approximation, the SH peak power follows:

where n _ω is the refractive index at the pulsation ω, l is the sample thickness, P _c,ω is the pump peak power, w₀ is the beam waist radius, L_c is the coherence length, θ_i is the internal propagation angle of the beam inside the sample, χ ⁽²⁾(x) is the induced nonlinear spatial distribution, and T(θ_i ) is a function which takes into account Fresnel transmission coefficients at the interfaces between the sample and the air [12].

 

Fig. 5. Experimental SH power recorded using MF method (full squares). The line stands for the SH power calculated by using the nonlinear spatial distribution shown in Fig.4

Download Full Size | PDF

The determination of the χ ⁽²⁾ magnitude is then performed by comparing the experimental SH peak power to the theoretical one, calculated using the profile of Fig. 4. The nonlinearity value is thus obtained for the best fit of the experimental MF data as shown in Fig. 5. The magnitude of the χ ⁽²⁾ susceptibility is finally estimated to be about 4.4±0.4 pm/V at maximum.

In order to reach a better understanding of the second-order NL optical properties of sulfide glasses, further studies are now in progress, such as the influence of the different elements composing the glass on the SHG level.

5. Conclusion

Second harmonic generation was studied in thermally poled Ga₅Ge₂₀Sb₁₀S₆₅ chalcogenide glass. Samples were poled under 4 kV applied voltage, at different temperatures during 30 minutes. For each temperature a SH signal was observed, but for temperatures higher than 230°C the samples present surface damages. An original method, based on the remaining SH signal measurement after NaOH etching was used to characterize the induced effect. To our best knowledge, it is the first time that such accuracy is achieved in chalcogenide NL profile characterization. For the sample poled at 230°C, a large NL second-order susceptibility χ ⁽²⁾ of about 4.4±0.4 pm/V was measured. These results open up the possibility of active devices fabrication in the IR domain.