1. Introduction

Silicon has been the prime candidate for optical integration in recent years. Using silicon as a platform for optical devices has the advantage of being compatible with complementary metal-oxide semiconductor (CMOS) nanofabrication techniques, which are well established in microelectronic industry. Because of its centrosymmetric crystalline structure, unstrained silicon, however, does not show any second order (χ⁽²⁾) non-linearity [1]. Hybrid silicon-organic devices, however, combine the mature silicon technology with the outstanding non-linear response of organic polymers which can be engineered by molecular design [2]. Electro-optic (EO) coefficients (r₃₃) of more than 300 pm/V have been reported for polymeric materials which is ten times more than in currently used materials like lithium niobate [3]. This has led to the realization of highly efficient optical devices such as optical interconnects [4], ultra-fast EO modulators [5,6], photonic sensors [7] and couplers [8]. In order to induce a macroscopic χ⁽²⁾ electro-optic response in polymers, centrosymmetry needs to be broken. This is commonly accomplished by orienting the dipolar molecules under a high electric field whilst heating them to a high mobility viscoelastic state close to their glass transition temperature [9]. The degree of poling induced noncentrosymmetrical order depends upon the strength of the poling field. Dielectric breakdown of the samples at high field strength imposes an upper limit for the maximally applicable poling voltage [10,11].

EO polymers are particularly interesting when infiltrated in slotted waveguides, since this configuration results in a strong overlap between the guided optical mode and the EO material. EO modulators with highly compact footprint, low driving voltage and outstanding speed results have been realized using this approach [6,12]. Despite the great successes, silicon - organic slot modulators have not yet realized their full potential. A series of studies using different types of polymers reports reduced in-device r₃₃ as compared to benchmark values obtained on parallel-plate poled thin film samples [13–18]. In silicon slot modulators, the polymers are poled inside the slot by applying the poling voltage across the two silicon rails alongside the slot that serve as electrodes. The reduced performance of EO polymers in such devices is usually ascribed to incomplete poling, due to geometric constraints in relation with surface anchoring in nanoscale slots that impede a full orientation of the molecules under the electric field. Considerable improvement of the in-device r₃₃ was achieved in widened slot geometries, which demonstrates that narrow slot dimensions can indeed circumvent the efficient poling in devices [19–21]. Recently, it has been shown that a binary chromophore organic glass enables higher in-device r₃₃, as the substance was found to sustain higher poling voltages when infiltrated in silicon slots [22].

While benchmark values for the r₃₃ performance of EO polymers, as well as the impact of dielectric coating layers on the poling process, have been investigated on transparent electrode materials such as ITO [23] the poling process on high refractive index silicon substrate has not yet been systematically studied. This is because the r₃₃ are usually probed by using a Teng-Man ellipsometric reflection technique [24] and a parallel-plate poled thin film sample configuration with transparent substrate materials [25,26]. However, this specific technique cannot be directly applied to EO polymers deposited onto silicon substrate, since silicon substrate, with a high refractive index, leads to multiple reflections in the substrate that impede deriving r₃₃ in this setup [27]. Therefore, the r₃₃ of EO polymers poled on silicon substrate can only be determined indirectly from the performance of the corresponding silicon-hybrid device and the poling process has not yet been studied under controlled and comparable conditions. Recently, an extension of the Teng-Man technique was published that allows one to independently measure both EO coefficients r₃₃ and r₁₃ on high refractive index substrates by taking into account the multiple reflections [27]. Here, we employ this technique to study the poling process of frequently used EO polymers (APC - CKL1) poled on silicon substrate, ITO substrate and Al₂O₃ coated silicon substrate in a thin film geometry under comparable conditions.

2. Methods

2.1 Sample preparation

Three different polymer thin film samples with different electrode materials were fabricated. The EO polymer we utilized is a frequently used guest host system with an EO active chromophore AJ-CKL1 doped 25 wt% into amorphous polycarbonate (APC) [28]. We spin coated polymer thin films from solution (APC/CKL1 in cyclopentanone (8 wt%)) at 2000 rpm onto three different types of substrates, namely: Double polished silicon substrates (n-type, doping concentration: 10-15/cm³, resistivity: 10 Ωcm), ITO coated glass substrates and silicon substrate coated with 5 nm Al₂O₃. The thickness of the polymer layer is d_poly ≈800 nm on all samples in order to provide comparability with other studies [23] and was ascertained using a Dektak profilometer. After polymer deposition, the samples were placed in an evacuated oven at a temperature of 90 °C over-night to remove any remaining solvent. The initially spin coated samples had a size of 9 cm². However, in the later sample processing only the central parts of these samples were used to ensure the polymer film has a homogenous coating thickness. Subsequently, circularly shaped gold electrodes of 5 mm diameter and 100 nm thickness were thermally evaporated on top of the polymer film. The electrodes were positioned in the middle of the sample where the spin coated polymer film is of homogeneous thickness. Thin film samples (TFSs) made from silicon substrate, ITO and Al₂O₃ coated silicon substrate are referred to as TFS_Si, TFS_ITO, TFS_{Si + Al2O3}, respectively, where the subscript denotes the different electrode materials. Sketches of the three different types of samples are shown in Fig. 1 (a)-(c). The deposition of the Al₂O₃layer onto the silicon substrate was carried out by atomic layer deposition [29]. The reaction cycle involved two steps using trimethylaluminum and water as precursors. The deposition process comprised 45 deposition cycles at a temperature of 200 °C.

 

Fig. 1 Schematic of the sample geometry for samples with different electrode materials. (a): ITO-glass electrode. (b): silicon electrode. (c): Al₂O₃ coated silicon electrode.

Download Full Size | PDF

2.2 Poling procedure

Prior to r₃₃ measurements, all samples were poled to induce EO activity in the polymer films. In detail, poling took place under nitrogen atmosphere by applying an electric field between the substrate and the gold electrode whilst heating the samples to 145 °C at a rate of 10 °C/minute. After holding the temperature at 145 °C for 1 minute, the samples were cooled rapidly, thus freezing the oriented state of the molecules. The applied electric field was kept constant during the procedure. The substrate material was biased positively in order to provide comparability to [23] where the poling of APC-CKL1 on ITO substrates and the effect of a metal oxide coating layer has been previously investigated.

2.3 r₃₃ measurements

r₃₃ measurements were conducted utilizing the recently published modified Teng-Man technique [27] that allows for directly probing the r₃₃ of EO polymers on silicon substrate by exploiting Fabry-Perot resonances in the high refractive index material. The optical setup is equipped with a tunable laser source and the experiment was carried out at a wavelength of 1350 nm.

2.4 I(V) characteristics

I(V) characteristics were acquired at a temperature of 140 °C - close to the glass transition temperature of the polymer. A Keithley 617 picoamperemeter was used to monitor the current through the sample whilst applying a voltage between the substrate and the gold electrode. For the reason of consistency, the applied voltage was increased at the same rate (0.2 V/s) in all measurements.

3. Results and discussion

Figure 2 shows r₃₃ values obtained for TFS_Si, TFS_ITO, TFS_{Si + Al2O3} for different poling fields (E_pol).

 

Fig. 2 r₃₃ for different electrode materials plotted versus the poling field (E_pol). Blue squares: r₃₃ of TFS_Si. Red triangles: r₃₃ of TFS_{Si + Al2O3}. Black circles: r₃₃ of TFS_ITO.

Download Full Size | PDF

We observe significantly different r₃₃ results for the three electrode materials for all poling voltages plotted in Fig. 2. For the ITO electrode, we find the highest r₃₃ values for all poling voltages. In particular, we attain a value of r₃₃ = 80 pm/V at E_pol = 100 V/μm and a poling efficiency r₃₃/ E_pol = 0.88 nm²/V² as deduced from the slope of the linear fit of r₃₃ vs. E_pol. These results conform to previous reports for this electrode material and EO chromophore used (r₃₃/ E_pol = 1.12 nm²/V²) taking into account the slightly different chromophore loading densities [23,28]. For the silicon substrate, we observe a significantly reduced poling efficiency of r₃₃/E_pol = 0.19 nm²/V² compared to the ITO substrate. In particular, we attain a value of r₃₃ = 29 pm/V at E_pol = 130 V. These findings may be partly related to the performance of EO polymers in other silicon-organic hybrid devices. We note that the poling efficiency on the silicon substrate is reduced compared to the ITO substrate in comparable and controlled thin film geometry at a relatively large polymer film thickness of 800 nm which is free from geometric constraints as imposed by nanoscaled slots in hybrid devices. This indicates that the often observed reduced performance of EO polymers infiltrated in silicon slots [13,14,16,18–20,30] may also be the consequence of a poor poling efficiency of EO polymers in combination with the silicon substrate and may not be solely explained by geometric constraints of narrow slots in relation with surface anchoring. Samples made from a silicon electrode coated with 5 nm Al₂O₃ (TFS_{Si + Al2O3}) were found to be the most robust under high applied fields allowing for higher applicable voltages as compared to the other sample types. The increased stability is due to the additional dielectric layer incorporated in TFS_{Si + Al2O3} which reduces the leak-through current during poling. Figure 3(a) and 3(b) shows the leak-through current during poling of TFS_Si and TFS_{Si + Al2O3}, respectively, plotted against the time along with the corresponding temperature.

 

Fig. 3 Current density and temperature vs. time in poling experiments at E_pol = 100 V/μm for TFS_Si (a) and TFS_{Si + Al2O3} (b). In the poling procedure the samples were heated to T = 145°C at a rate of 10°C/min and subsequently cooled down. The electric field remained constant during the process.

Download Full Size | PDF

The suppression of the current efficiently reduces the probability of singular dielectric breakdown events in the samples at high electric fields. Film defects and inclusions, such as voids and impurities, lead to local fluctuations of the conductivity which result in current hot spots that induce breakdown of the samples. The effect of defects is additionally enhanced by local heating in association with higher currents which further increases the conductivity in the samples. We assume that such current hot spots form randomly distributed underneath the electrodes [10,11]. However, we cannot exclude that the breakdown events occur mainly at the edge of the electrode where the electric field is expected to be enhanced which may lead to locally enhanced currents that additionally increase the probability for breakdown. The reduction of the current by the Al₂O₃ layer causes detrimental hot spots to form at higher poling voltages. Therefore, the reduction of the current shifts the limit for catastrophic sample breakdown to higher poling fields, which permit a higher orientational order in the film and, consequently, improved r₃₃ [10,11]. The best achieved value of r₃₃ for TFS_{Si + Al2O3} is r₃₃ = 55 pm/V at E_pol = 200 V/μm which is a roughly twofold enhancement compared to TFS_Si (r₃₃ = 29 pm/V at E_pol = 130 V/μm). We wish to point out that the dielectric layer in TFS_{Si + Al2O3} only has a thickness of 5 nm which makes it suitable for integration in silicon-organic hybrid devices without increasing losses or diminishing high quality optical guiding properties. We note that this method significantly improves the attainable absolute value of r₃₃ on silicon substrates, however, the reduced poling efficiency on silicon substrate remains unaffected. For TFS_{Si + Al2O3} we find a poling efficiency of r₃₃/E_pol = 0.22 nm²/V² which represents no significant improvement compared to TFS_Si (r₃₃/E_pol = 0.19 nm²/V²). This demonstrates that the poling efficiency is independent of the leak through current during the poling process. The Al₂O₃ layer in TFS_{Si + Al2O3} reduces the current density but does not reduce the poling field dropping over the polymer film since we observe comparable r₃₃ results for TFS_Si and TFS_{Si + Al2O3}. Therefore, the role of the Al₂O₃ modified interface in TFS_{Si + Al2O3} must be suppressing charge injection at the silicon polymer interface without reducing the poling voltage dropping over the polymer film. A similar effect of a metal oxide barrier layer was observed by Huang et al. [23] who investigated the impact of a metal oxide layer deposited on ITO substrates.

In order to investigate the origin for the reduced poling efficiency on silicon substrate we examine the electronic conduction process in TFS_ITO and TFS_Si. Figure 4 shows I(V) characteristics for TFS_ITO and TFS_Si plotted logarithmically versus the square root of the applied electric field.

 

Fig. 4 Current density as a function of the square root of the applied electric field for TFS_Si (blue) and TFS_ITO (black). All data was acquired at a temperature of 140°C – close to the glass transition temperature of the doped polymer.

Download Full Size | PDF

For both samples, we find a deviation from simple ohmic behavior. For moderate voltages we find straight lines indicating a $j~e^{c\sqrt{E}}$ field dependence of the current density j, with c being a constant and E being the applied electric field. A $j~e^{c\sqrt{E}}$ dependence has been frequently reported for EO polymers [10,11,31] and is indicative for either a Pool-Frenkel type conduction process or Schottky thermionic charge emission over a potential barrier lowered by the image force potential of the electrons [11]. As the two processes show similar field dependences, they often cannot be directly distinguished [32].

For TFS_Si, we observe a transition for high applied fields from the $j~e^{c\sqrt{E}}$ field dependence into a sub-ohmic current with $j~E^{\gamma },\gamma <1$. For even higher fields (data not shown) we either observe dielectric breakdown of the samples or significant physical degradation in the samples that results in an irreversible reduction of the current. However, the transition to sub-ohmic behavior for TFS_Si can be reproduced in any sample by repeated runs. Therefore, the sub-ohmic characteristics are not related to any physical degradation of the EO polymer observed at very high currents or voltages. Furthermore, our calculations show that the ohmic bulk resistance of the silicon electrode does not limit the conduction of the sample in this field range and the applied voltage will thus fully drop over the polymer film. The sensitivity of the electronic conduction in the different samples towards the electrode material indicates an interface controlled nature of the conduction process [11].

A saturation current with a sub-ohmic characteristic is well known for metal semiconductor contacts made of, for example, gold and semi-insulating GaAs [33–35]. In this case, the sub-ohmic behavior is the consequence of thermionic charge emission across a Schottky potential barrier at an inversely biased metal-semiconductor junction [33]. Under the assumption of a purely thermionic current transport, without the effect of barrier lowering, the current density is expected to be independent on the applied voltage [33,36]. We consider the sub-ohmic current observed for TFS_Si as an indicator for a similar Schottky diode-like effect in the sample. We suppose that the sub-ohmic current in TFS_Si is due to a potential barrier at the silicon-organic interface, acting similar to the potential barrier in schottky diodes, giving rise to an interface controlled thermionic emission current with sub-ohmic characteristic. The potential barrier is accompanied by a space-charge zone in the polymer which effectively acts as a semiconductor. We hypothesize that this space charge zone causes the reduced poling efficiency observed for polymers poled on silicon substrates. A space-charge zone in the polymer film close to the electrode interface will distort the field distribution of the poling field in the polymer film during poling [23]. As was also shown by Huang et al. [23] inhomogeneities in the electric field distribution in the polymer film are detrimental for the poling efficiency in that they may lead to partial discharge in the film even if the average applied field is lower than the critical breakdown voltage of the system. Additionally, a non-uniform electric field could act as dielectrophoretic force on polarizable particles which may result in microphase separation which impedes the efficient poling under high field strength [23]. We note that the used guest host polymers have a relatively high loading density of highly polarizable chromophores and it is therefore expected that the system reacts very sensitively towards an inhomogeneous internal electric field. Our data suggest that the effect of field distortion in the polymer film at the electrode interface is stronger for the silicon electrode compared to the ITO electrode, which does not lead to a comparable formation of a space-charge zone with sub-ohmic current characteristics. We therefore observe a reduced poling efficiency on the silicon substrate as compared to ITO electrodes.

4. Conclusion

We presented a thin film study of the poling induced electro-optic response of APC-CKL1 on different electrode materials and the underlying conduction properties. We find a reduced poling efficiency of APC-CKL1 on silicon substrates, which we attribute to space charge formation at the silicon-organic interface that causes a distorted field distribution in the polymer film close to the electrode, which reduces the poling efficiency. Our data suggest that the frequently reported reduced in-device performance of polymers in silicon-organic hybrid devices may not be entirely explained by geometric constraints imposed by nanoscale device dimensions, but may also be the consequence of a poor poling efficiency of EO polymers in combination with silicon substrate. By employing an ultra-thin layer of Al₂O₃ we demonstrate a method to significantly enhance the maximally applicable poling voltage to silicon electrode samples which is a viable route to roughly double the r₃₃ in silicon-organic hybrid devices. However, the poling efficiency does not significantly improve compared to the case of untreated silicon electrodes. Subsequent studies will have to clarify whether these conclusions also hold for other EO polymers such as APC-YLD 124 [16], PMMA-AJSP100 [17] and PMMA-ALJZ53 [18] for which reduced EO activity in silicon hybrid devices was also reported. Additionally the effect of p-type silicon substrate, as opposed to the n-type substrate used in this work, and doping concentration will have to be investigated. In the case of binary chromophore organic glass [22] as well as SEO125 [21], exceptional high r₃₃ values were reported in silicon hybrid devices which suggest that the silicon substrate has no detrimental effect on these materials. Future experiments should also investigate the impact of the native oxide layer on the silicon substrate (10-30 Å) which may potentially affect the interface properties and in turn the poling efficiency. The strong sensitivity of the poling efficiency of APC-CKL1 on the electrode material stresses the importance of considering the compatibility with silicon substrate as a design parameter in EO polymer research. It also demonstrates the need for extensive electronic modelling of the silicon-polymer interface and, consequently, the establishment of electronic matching techniques for the efficient application of EO polymers in silicon hybrid devices such as optical interconnects [4], EO modulators [5], photonic sensors [7] and couplers [8].

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. K.M.S. and S.P. contributed equally to this work.