1. INTRODUCTION

The recent advantages in the field of materials engineering have allowed the development of a completely new class of materials with properties unparalleled in nature. Due to their artificial complexity, the properties of metamaterials can be designed to a large extent. One of the main routes of the development of such structures utilizes the collection of individual particles or atomic mono-layers of metallic surfaces [1–4]. These are in turn assembled into nanocomposites with effective properties, which often are dramatically different from the nanocomposite constituents [5,6]. In recent years, it has been proposed to broaden the palette of plasmonic materials with transparent conducting oxides (TCOs) or semiconductors [7]. Due to this remarkable elasticity in the engineering of metamaterials, they are finding now broad applications, including: negative refraction [8,9], optical cloaking [10], biosensing [11,12], etc. The same diversity of possible materials and techniques contributes to the already demanding task of modeling of such structures. In this study, we consider metamaterials for possible nonlinear optics applications. From this point of view, the effective medium approach is particularly important, because the macroscopic properties of the sample can play a crucial role in determining whether a significant response can be obtained. In the case of second-order nonlinear processes, the requirements for significant nonlinear response are very demanding. First of all, second-order effects, e.g., second-harmonic generation (SHG), are possible only in noncentrosymmetric media. To facilitate such a process, solutions based on multilayer systems have been proposed. In such composites the symmetry breaking effect is achieved as a result of folding thin layers of different materials, or by simply changing the properties of the same material by doping it. The gigantic enhancement of the second-order nonlinear effects was observed in structures consisting of silica glass that underwent the thermal poling process [13]. A similar amplification of the response was observed in a composite of non-doped and doped with a phosphorus donor silica glass [14]. In recent years, reports have emerged in which a group of three different materials form a two-dimensional ABC-type crystal. In this material, three dielectric materials are stacked in such a manner that the resulting metamaterial is noncentrosymmetric in the direction of the deposition, therefore giving rise to the second-order nonlinear response [15,16]. Moreover, roughness of the metallic surface enhances the SHG responses [17], and the composites made of alternating layers of silver-decorated silica glass nanoparticles and fused silica glass give rise to the SHG response by the increase in an overall noncentrosymmetry. That in the end leads to a SHG that scales with the number of layers of the nanocomposite elementary cell [18]. In this study, the effect of induced symmetry breaking on the semiconductor–dielectric interface where an additional symmetry-breaking mechanism is activated by the temperature is presented.

2. SAMPLES AND MEASUREMENTS

The samples were prepared in the clean-room laboratory of the Institute of Microelectronics and Optoelectronics of Warsaw University of Technology. The 50 nm layer of silicon nitride (SiN) was deposited utilizing plasma enhanced chemical vapor deposition (PECVD) [19]. The 10 nm thick layer of indium gallium zinc oxide (IGZO) was deposited by a pulsed-DC magnetron sputtering (MS) process [20]. Depending on the assumed order of layers in the elementary cell, the first material (SiN or IGZO) is applied on a quartz glass substrate (SILUX), and then the other material is applied directly on top, completing the bi-layer. The two-stage fabrication process is then repeated up to four times, therefore creating the set of samples with a varying number of the elementary cells ${N}$. Moreover, for the sake of thickness verification, the deposited materials and periodic structures were formed on reference Si substrates. The thickness of obtained materials was identified with spectroscopic ellipsometry (Horiba Jobin-Yvon UVISEL) in the range of 250 to 850 nm. A schematic representation of prepared samples with SiN-IGZO bi-layers is depicted in Fig. 1(a). The samples with the order of application IGZO-SiN are presented in Fig. 1(b). The surface area of the samples is very large, up to $5\,\,{\rm cm^2}$, and it is limited mainly by the surface area of the substrate. The deposited layers are uniform across the sample, which is confirmed by performing multiple measurements across the sample surface, and the results are consistent with the average over many measurements, especially for the samples with a larger number of layers. To separate the surface and bulk effects, a set of two control samples was prepared. The schematic cross-sections of both control samples C1 and C2 are presented in Figs. 1(c) and 1(d), respectively. Instead of a multi-stage process, a 200 nm layer of SiN and a 40 nm layer of IGZO were simply fabricated on a SILUX quartz glass plate in the correct order, therefore creating one single bi-layer with a thickness equal to the structured samples with the number of layers ${N} = {4}$. All the Maker-fringe SHG experiment results are scaled with respect to the maximum value of the SHG measured for the pure SILUX quartz glass plate. The fabricated samples were investigated in the setup presented in Fig. 2. The pulsed laser NKT Origami 10XP produces 320 fs pulses with the energy of $10\,\,\unicode{x00B5}\rm J$ and a repetition rate of 50 kHZ at 1030 nm wavelength. The average power, measured right before the sample, was 350 mW. The light from the laser was weakly focused on a sample with a planoconvex lens of a 50 cm focal length, resulting in a focal spot with a size of about $400\,\,\unicode{x00B5}\rm m$. The input polarization state was controlled with a Glan–Taylor polarizer and a half-wave plate, and set to be parallel to the plane of incidence ($p$) initially. The second-harmonic signal generated on the optical elements before the sample was cut out with a filter with a cut-on wavelength of 1000 nm. The second-harmonic signal generated on a sample was measured with a photon multiplier tube. The light of the laser was cut out right after the sample by a block filter with a cut-off wavelength of 1000 nm, and by a laser-line bandpass filter with a center wavelength at 520 nm and FWHM of 40 nm. All the elements give absolute certainty that the measured signal comes from the sample itself. During the measurement, the angle of incidence is varied continuously from 0° to 90° with respect to the sample surface, and the result is recorded for every degree of rotation, giving the Maker-fringe pattern [21]. In another experiment, the character of the nonlinear response was determined by setting the incidence angle to a value corresponding to the strongest SHG response, and rotating the polarization direction of the incoming light. Such a measurement gives a hint about the character of the nonlinear interaction between the light and a structure.

 

Fig. 1. Schematic representation of the measured structures: (a) bi-layers of 50 nm SiN and 10 nm IGZO; (b) bi-layers of 10 nm IGZO and 50 nm SiN. In both cases, the set consisting of ${N} = [1-4]$ bi-layers was prepared. (c) Control sample C1 and (d) control sample C2.

Download Full Size | PDF

 

Fig. 2. Measurement setup used in all experiments.

Download Full Size | PDF

3. RESULTS

First, the response of the samples presented schematically in Fig. 1(a) is measured. The full measurement cycle is made with the set of four samples with the number of bi-layers from ${N} = {1}$ to ${N} = {4}$. A control sample C1 [Fig. 1(c)] contains the same amounts of deposited materials as in the case of the structured sample with the number of layers ${N} = {4}$, but without the layered structure. Therefore, a sample with the 200 nm layer of SiN applied to the quartz glass substrate with a 40 nm layer of IGZO added on top of the SiN was fabricated. The role of a control sample is to investigate the mechanism of the second-harmonic radiation and its nature. All the measurement results are presented in Fig. 3(a), and it is visible that the response rises together with the increasing number of layers. The most interesting result is that among a pure substrate, a sample with the number of SiN-IGZO layers ${N} = {1}$, and a control sample C1 [Fig. 3(b)], there is not much difference in the generated SHG, especially in comparison with the samples with a higher number of bi-layers, where the increase in generated SHG exceeds two orders of magnitude for ${N} = {4}$. One might argue that the main driving mechanism of the SHG increase is related to the order of the application of the SiN and IGZO layers in a single bi-layer. Motivated by the above, the samples represented schematically in Fig. 1(b) were fabricated. Effectively, both sets of samples are similar in that they have equal amounts of the applied materials. The only difference is that the layer of IGZO is applied on a substrate first, and then the layer of SiN is applied on top. The new set of samples is also supplemented with a control sample C2 [Fig. 1(d)], which is identical to C1, except for the order of application of the layers. The SHG response of the resulting set of four samples with the varying number of bi-layers (from ${N} = {1}$ to ${N} = {4}$) was then measured. As demonstrated in Fig. 3(c), the rise in the SHG signal from a single bi-layer is now substantial, more than one order of magnitude higher than the signal from a clean substrate. Moreover, a comparison between the control sample C2 (total thickness 240 nm) and a single IGZO-SiN bi-layer (total thickness 60 nm) leads to a very interesting observation. An effectively thinner sample made with the same materials as control sample C2 gives surprisingly close to four times a stronger response [Fig. 3(d)], therefore leading to the conclusion that surface effects dominate the nonlinear response of the prepared samples.

 

Fig. 3. Second-harmonic response of the samples with: (a) SiN-IGZO sandwich with the number of bi-layers from ${N} = {1}$ to ${N} = {4}$; (b) comparison of ${N} = {1}$ to control sample C1 and substrate; (c) samples with IGZO-SiN sandwich with the number of bi-layers from ${N} = {1}$ to ${N} = {4}$; (d) comparison of ${N} = {1}$ to control sample C2 and substrate. Multiple measurements across the sample surface were performed, to account for possible inconsistencies in the sample structure. However, the results proved to be consistent with the average over many measurements, which is illustrated in (c), where the series of nine measurements across the sample surface is presented for the sample with ${N} = {4}$ bi-layers.

Download Full Size | PDF

4. DISCUSSION

Comparing the results of all the measurements leads to the conclusion that the character of the response is related strongly to a combination of the fabrication techniques used, alongside with the order of the application of the layers in our samples. Having measured effectively identical media, which consist of three layers (1 mm quartz glass, 50 nm SiN, 10 nm IGZO), where the order of SiN and IGZO layers was exchanged, about a 25-fold increase in the measured SHG signal, in the case in which IGZO was applied on the substrate as a first, was observed. Moreover, if the combination of two SiN-IGZO bi-layers is considered, it effectively translates into one IGZO-SiN bi-layer; and what follows is that for every number ${N}$ of SiN-IGZO bi-layers, there is a number (${N} - {1}$) of IGZO-SiN bi-layers in the structure. Therefore, a direct comparison of the measurement results for the samples presented in Figs. 1(a) and 1(b) can be done. It can be observed that for the equal effective number of IGZO-SiN bi-layers, we get a very good agreement between measured Maker-fringe curves for both sets of samples (Fig. 4). Since one can be sure that the nonlinear effect is driven mainly by the sequence of the fabrication process, some drastic changes in the structure of the prepared samples can be expected due to the fabrication process. To address the raised assumption, examinations using atomic force microscopy (AFM) of the single deposited layers together with the complete bi-layers combined in the elementary cell of our samples were performed. The results, presented in Fig. 5, show a massive increase of the surface roughness in the case of using the PECVD technique after application of IGZO with the MS. The root mean square (rms) parameter for the prepared samples is maintained below 0.5 nm for all cases in which the PECVD technique is used before MS or not used at all. However, if the layer of SiN is applied with PECVD on top of the IGZO layer, the topography of the surface changes dramatically, and the roughness parameter increases to the value of ${\rm rms} = {41.2}\;{\rm nm}$. Such a dramatic effect can be explained by the large difference in the temperature for both fabrication methods (room temperature for MS versus $300^\circ \rm C$ for PECVD). The temperature of the PECVD process gives rise to structural changes in the IGZO layer, which contribute to the overall roughness of the bi-layer. Such structural changes have already been reported, and it was shown that the optical and electric properties of IGZO layers can be controlled to an extent by changing the substrate temperature or oxygen pressure during the process of fabrication [22]. In our work, structural changes in IGZO are induced by the PECVD process, and in Fig. 5(d), small polycrystalline domains arising because of the comparably high temperature of the PECVD process can be observed. Due to a phase transition between the amorphous and polycrystalline phase of IGZO, one could expect the rise of anisotropy in the nonlinear response of the structure. However, the polycrystalline domains are considerably smaller than the size of the focal spot for the lens used in the experiments, so the character of the IGZO layer still should stay isotropic. It was confirmed by means of a simple measurement of the SHG signals for different combinations of the input–output polarization configurations. The results presented in Table 1 show that the maximum SHG is observed for the input–output polarization combination parallel to the plane of incidence ($p_{{\rm in}}-p_{{\rm out}}$). In comparison, very weak SHG signals measured for the polarization combinations $s_{{\rm in}}-s_{{\rm out}}$ and $p_{{\rm in}}-s_{{\rm out}}$ confirm that the samples are indeed isotropic in the plane of the sample.

 

Fig. 4. Comparison of the Maker-fringe curves from the samples in Figs. 1(a) and (b) with the same number of effective IGZO-SiN layers.

Download Full Size | PDF

 

Fig. 5. Atomic force microscope images. (a) Surface of 50 nm of SiN on a silicon substrate, with root mean square roughness parameter ${\rm rms} = {0.27}\;{\rm nm}$. (b) Surface of 10 nm IGZO on a silicon substrate with ${\rm rms} = {0.37}\;{\rm nm}$. (c) Surface of the SiN-IGZO bi-layer (50 nm, 10 nm, respectively) with ${\rm rms} = {0.49}\;{\rm nm}$. (d) Surface of the IGZO-SiN bi-layer (10 nm, 50 nm, respectively) with ${\rm rms} = {41.2}\;{\rm nm}$.

Download Full Size | PDF

Visible Maker-fringes are the result of the interference between the SHG signals and the bottom side of the 1 mm glass plate. The Maker-fringe effect for the structure itself cannot be resolved, as the thickest structure has features lower than the wavelength of the SHG. The differences in the susceptibility components between the active layer and the bottom side of the substrate give rise to the shift of the SHG maximum between different samples. In the case of all the performed measurements, the maximum value of the SHG signal occurs for an incidence angle between $60^\circ$ and $70^ \circ$. Most importantly, the SHG strength increases with the number of bi-layers for all the measured samples, and the introduced structure is the main source of the strong rise in SHG signals. This result points to the fact that the surface effects play a central role in enhancing SHG—more precisely, SHG arises from the multiple coherent second-harmonic sources induced by symmetry breaking arising mainly due to a thermal effect of structural change in the layers of IGZO. The value of conversion efficiency was estimated to be on the order of ${10^{- 6}}$ for the four-layer IGZO-SiN sample, which is an excellent result for a structure of only 240 nm of thickness.

Table 1. Normalized SHG Signals for Different Input–Output Polarization Configurations

View Table

The proposed structure is not yet optimized, and it will be further modified to accommodate optical nonlinear applications. The periodic structure, in principle, permits to choose the constituents in a way to allow for controlled modification of the effective permittivity of the resulting medium. The task, however, is not that straightforward due to the very exotic nature of the thin layers, as the material parameters of such layers are not easy to determine, especially when the thickness of the examined structures is up to a few tens of nanometers. In future work, the introduction of a metallic layer to the periodic structure and simplification of the fabrication process are planned. The intended aim is a structure with engineered effective permittivity that can be modified by the polarizing voltage applied to it.

5. CONCLUSION

In conclusion, investigations of the SHG response of multilayer structures, containing ultra-thin films of dielectric and semiconductor material, were presented. The origin of the second-order nonlinear response is the symmetry breaking between consecutive layers of the structure. Moreover, this effect is altered by the technological process of sample fabrication, in which the thermal effects of the structural change in the layer of a semiconductor play a crucial role. Due to these effects, the order in which layers are applied determines the roughness of the final bi-layer, therefore resulting in an additional symmetry-breaking mechanism. Obtained results were examined with a set of control samples that provided definite proof of the role of the introduced structure to the nonlinear response. The utilized fabrication process is very straightforward, and we expect to further develop the proposed structures and optimize them for the nonlinear optics applications.