1. Introduction

Chalcogenide glasses are very promising materials for all-optical processing in fiber optics and integrated optics since they have a good transparency in the infrared region and a high third-order optical nonlinearity [1,2]. In particular, As₂Se₃ and Ag-As₂Se₃ (i.e., Ag_x(As_0.4 Se_0.6)_100-x) glasses posses a high nonlinearity greater than 1000 times that of fused silica [3,4] and an ultrafast time response because of non-resonant Kerr nonlinearity. Recently, it has also been found that these glasses have large Raman gains nearly 800 times that of silica [5] and large Brillouin gains nearly 20 times that of silica [6]. These large optical nonlinearities could be used for all-optical devices such as bistable optical devices and optical switches, parametric amplifiers, Raman (or Brillouin) amplifiers and Raman (or Brillouin) lasers. Moreover, it has recently been found that the addition of Ag into As₂Se₃ glasses eliminates or weakens a harmful photo-oxidation of the glass surface, i.e., Ag-As₂Se₃ glasses are stable with respect to exposure to band-gap light, air, and moisture [7].

To realize all-optical devices with As-Se or Ag-As-Se chalcogenide glasses, we must be able to fabricate a three-dimensional waveguide with the same materials at least. In this case, the single-mode operation and low propagation loss are required for ensuring the adequate operation of the desired devices. Propagation loss in a straight optical waveguide is generally attributable to absorption and scattering. Absorption loss in an amorphous film is caused by electronic band-gap excitations, the Rayleigh scattering, the electronic transitions of impurity, absorption from impurity molecular vibrations, and IR absorption from multiphonon. Attenuation on the short-wavelength side of the optical window is dominated by electronic band-gap excitations and the rest of the sources dominate attenuation within the window. On the other hand, scattering loss in an optical waveguide is caused by a variety of structural imperfections of the waveguide. The most dominant scattering loss is surface scattering loss, which can be decreased by reducing the interaction between the propagating wave and the surface of the waveguide. Moreover, the cross section of the waveguide becomes small to ensure its single-mode operation since the refractive index of As-Se or Ag-As-Se glasses is high (∼2.5-∼3.2). Although several types of three-dimensional waveguides have been devised to date, a strip-loaded waveguide is the most adequate among them because of low scattering loss, easiness of fabrication, large cross section, and good controllability of propagation properties.

Regarding optical waveguides made of As-Se or Ag-As-Se glasses, DeCorby et al. [8–10] have recently proposed a novel fabrication process using photodarkening and subsequent selective wet etching, and have reported the propagation properties of the fabricated optical waveguides. They have fabricated As₂Se₃ shallow rib waveguides with a propagation loss as low as ∼0.25 dB/cm at a wavelength of 1.55 μm. Although they fabricated multimode Ag-doped strip waveguides too, the propagation loss was ∼9 dB/cm at 1.48 μm. Since experimental research on such optical waveguides using As-Se or Ag-As-Se glasses has just been begun, further investigation is required.

In this paper, we describe the experimental data on absorption loss in As₂Se₃ and Ag-As₂Se₃ bulk glasses and report the fabrication of low-loss single-mode strip-loaded waveguides using these glasses. The fabrication method used here is based on the standard photolithography, which can be applied to the fabrication of all-optical devices. In this work, Ag-As₂Se₃ glass with an Ag content of 3 at.% and pure Ag-As₂Se₃ glass were used as a guiding layer and a loading strip, respectively. The propagation loss was measured to be 0.5 dB/cm at a wavelength of 1.053 μm, which was attributed to absorption in the basic materials. If the waveguides are operated at the optical telecommunication wavelengths of 1.3 and 1.55 μm, we can expect lower losses. We conclude that the strip-loaded waveguide studied in this paper is suitable for all-optical device applications because of a high nonlinearity, low loss, and free photo-oxidation reaction.

2. Experiment

2.1 Absorption Loss in Bulk Glasses

Absorption loss in base materials must be investigated since it determines an attainable lower limit of the propagation loss in straight optical waveguides. Although we tried to estimate the absorption loss in evaporated films on a microscopic slide from the transmission spectrum using Swanepoel’s method [11], we could not determine it in the low loss region less than ∼100 cm^-1 because the thickness of the samples used for measurements was on the order of 1 μm. Therefore, we measured the absorption loss in Ag-As₂Se₃ bulk glasses from their transmission spectrum. The bulk glasses used in this measurement were prepared by a conventional melt-quenching method [12]. First, As₂Se₃ glasses were synthesized by melting the mixture of As and Se elements of 6N in an evacuated fused-quartz ampoule. The ampoule was placed in a rocking furnace at about 1000 °C for 25 h to increase the mixing and homogenization of the melt and then rapidly cooled by putting it into water. Next, Ag-As₂Se₃ glasses were prepared by heating the mixture of the As₂Se₃ glass and appropriate amounts of Ag of 5N in the same way. The samples used for measurements were obtained by cutting the glasses into parallel plates of 1-mm thickness and polishing both surfaces of the plate. The absorption coefficient α is determined from

where d is the thickness of the glass plate, T and R are the transmission coefficient of the plate and the reflection coefficient at the plate surface, respectively. For measurements of low absorption less than ∼1 cm^-1, a precise experiment of transmission and reflection spectrum of thicker samples is required. Therefore, we estimated the dependence of the absorption coefficient α on the wavelength by assuming that the absorption is zero (α=0) at a wavelength where the measured transmittance has a maximum T_m =(1-R_m )² in the near infrared region. That is, we calculated the absorption coefficient from α=(1/d)ln[1/(T/T_m )], where T/T_m is the normalized transmittance.

Figure 1 shows the dependence of the absorption coefficient α of Ag_x(As_0.4Se_0.6)_100-x bulk glasses on the wavelength for different values of the Ag content x. The Ag element acts as glass structure modifier in the low Ag content region (x=1-4 at.%) and as glass structure former in the high Ag content region (x=15-30 at.%) [13]. It should be noted that we cannot continuously change the Ag content x since the present ternary Ag-As-Se system has two separated glass-forming regions. On the other hand, we can prepare Ag-As₂Se₃ films with an arbitrary Ag content from 0 to 21 at.% by photodoping [14]. The absorption loss in dB/cm in bulk glasses is given by 4.343α, where α is expressed in cm^-1. Since we will make an experiment at a wavelength of 1.053 μm, we try to estimate the absorption coefficient in Ag_x(As_0.4Se_0.6)_100-x bulk glasses using Eq. (1) at that wavelength. From the measurement of Brewster’s angle, we have measured the refractive index of Ag_x(As_0.4Se_0.6)_100-x bulk glasses at 1.053 μm. The obtained refractive index was 2.81 and 2.92 for the Ag content x=0 and 4 at.%, respectively. Calculating Eq. (1) using the transmittance R given from the refractive index, we have the absorption loss α= 1.44 and 3.74 dB/cm for the Ag content x=0 and 4 at.%, respectively. These absorption losses agree well with values calculated using the normalized transmittance. We can conclude that the absorption loss in Ag-As₂Se₃ bulk glasses is 1-4 dB/cm for low Ag content (x=0-4 at.%) and is more than ∼20 dB/cm for high Ag content (x≥ 15at.%). Therefore, we use the Ag-As₂Se₃ glasses with low Ag content as the waveguide material in the present experiment.

 

Fig. 1. Absorption coefficient α of Ag_x(As_0.4Se_0.6)_100-x bulk glasses as a function of wavelength λ. (a) Ag content x=0-4 at.% , (b) x=15-30 at.%. The absorption loss in dB/cm is given by 4.343 α

Download Full Size | PDF

2.2 Fabrication of Waveguide

We fabricated a strip-loaded waveguide from As₂Se₃ and Ag-doped As₂Se₃ glasses using standard photolithographic techniques. The strip-loaded waveguide can be formed by loading a strip of dielectric material of less index, n_l , on top of a planar waveguide with refractive index n_f as shown in Fig. 2. The presence of the loading strip on top of the guiding layer makes the effective index in the region beneath it larger than that in the adjacent regions. In the case of the strip-loaded waveguide, we can easily control its propagation properties by changing the width w, height h, and refractive index n_l of the loading strip. It has well been recognized that the strip-loaded waveguide can be fabricated with relaxed requirements for resolution and edge roughness. Compared with the fabrication of a rib waveguide (where n_l =n_f ), we need not etch the lateral walls of a channel waveguide into a previously formed planar waveguide structure. From these features, we selected such a waveguide as a three-dimensional waveguide to be used in our experiment.

 

Fig. 2. Fabrication process of the strip-loaded waveguides and its cross section.

Download Full Size | PDF

We designed the single-mode strip-loaded waveguide using the effective index method and finite-element method [15]. We used PMMA (n_s =1.48), Ag-As₂Se₃ (n_f =2.595), and As₂Se₃ (n_l =2.523) glasses as an under-cladding layer, guiding layer, and loading strip, respectively. These three refractive indices are ones at a wavelength of 1.053 μm and the last two values were determined from the transmission spectrum of Ag-As₂Se₃ and As₂Se₃ films deposited on a microscopic slide using Swanepoel’s method [11]. The strip-loaded waveguide generally support two types of mode classified ${\mathrm{E}}_{\text{pq}}^{\mathrm{x}}$ (TM-like modes) and ${\mathrm{E}}_{\text{pq}}^{\mathrm{y}}$ (TE-like modes), where the subscripts p and q indicate, respectively, the number of extrema of the electric field in the x and y directions. For the ${\mathrm{E}}_{11}^{\mathrm{x}}$ mode (TM-like mode), the single-mode condition in the x direction is 0.75 μm<d<1.60 μm for a fixed strip height h=0.1 μm and the single-mode condition in the y direction is 0<w<7.36 μm if we select d=1.2 μm. On the other hand, for the ${\mathrm{E}}_{11}^{\mathrm{y}}$ modes (TE-like mode), the single-mode condition in the x direction is 0.64 μm<d<1.50 μm for h=0.1 μm and the single-mode condition in the y direction is 0<w<8.48 μm for a given value of the guiding layer thickness d=1.2 μm. It should be noted that the fundamental ${\mathrm{E}}_{11}^{\mathrm{x}}$ mode is not a truly guided mode but a leaky mode owing to TE↔TM mode coupling at the sides of strip [16,17]. On the other hand, the ${\mathrm{E}}_{11}^{\mathrm{y}}$ mode never becomes the leaky mode.

We describe the fabrication of the strip-loaded waveguides. First, a PMMA (electronic beam photoresist, OEBR-1000, Tokyo Ohka) film was deposited on a cleaned Si wafer as an under-cladding layer by spincoating and was prebaked at 170 °C for 30 min. The thickness of the under-cladding layer is sufficient to be 1 μm at least since the refractive index difference n_f - n_s is large. Next, we deposited a Ag-As₂Se₃ film on the under-cladding layer as a guiding layer. Since one cannot directly prepare Ag-As₂Se₃ films out of original ternary glasses by the evaporation method because of their phase separation, we doped Ag into an As₂Se₃ binary film by the photo-induced diffusion of metals [14]; An As₂Se₃ film was first deposited on the under-cladding layer by 1.2 μm at a rate of 3 nm/s and an Ag film was then evaporated on the top of the host film. In a preliminary experiment for gaining information on an increase in refractive index due to the Ag addition, a microscopic slide was used as a substrate instead of a Si wafer. For the As₂Se₃ film of 1.2 μm thickness, the Ag deposition of 20 nm-thickness leaded to an Ag content of 3.0 at.% and the index increase of 0.072. Finally, a loading strip was fabricated on top of the Ag-As₂Se₃ film by the standard lift-off method; A layer of photoresist (OEBR-800, Tokyo Ohka) was spincoated on the Ag-doped As₂Se₃ film and was baked at 80 °C for 30 min. The photoresist was exposed to UV light through a contact photomask that defines the shape of loading strip and was then developed to form a waveguide pattern. An As₂Se₃ film was deposited by 0.1 μm as a loading strip. The unnecessary As₂Se₃ film and photoresist in unmasked regions were removed with acetone to form a strip-loaded waveguide.

 

Fig. 3. SEM images of (a) the end-face of a cleaved strip-loaded waveguide and (b) top view of the same waveguide.

Download Full Size | PDF

Figure 3 shows scanning electron microscope (SEM) pictures of the fabricated strip-loaded waveguide with the strip width w=3.0 μm and the strip height h=0.1 μm. The SEM images show that the waveguide has been fabricated as designed and that each surface of the waveguide is smooth. This hints at the potential of the strip-loaded waveguide in achieving low-loss optical propagation.

2.3 Waveguide Characterizations

We simultaneously fabricated three sets of waveguides with two strip widths w=3.0 and 6.0 μm on a Si wafer to evaluate the strip-loaded waveguides. The fabricated waveguides were cleaved using a diamond cutter to obtain good quality end-faces. We here confirm the single-mode operation of the fabricated waveguides and measure the propagation loss using a Q-switched YLF laser (operating wavelength: 1.053 μm, pulse width: 23 ns, and repetition rate: 1 kHz).

The laser light was end-fire coupled onto the waveguide with a 40X microscope objective (NA=0.65) and the scattered and transmitted light from the waveguide was observed with a sensitive IR camera. Since the ${\mathrm{E}}_{\text{pq}}^{\mathrm{y}}$ modes are leaky modes and leakage loss could be added to the propagation loss in the waveguide, we propagated the ${\mathrm{E}}_{\text{pq}}^{\mathrm{x}}$ modes inside the waveguide by focusing the laser beam linearly polarized in the y direction on the waveguide end-face. The average input power to the waveguide was 270 μW.

First, we observed a characteristic streak of light scattered from the waveguide and the output light from the waveguide end-face. Figure 4 shows the scattered light streak from a strip-loaded waveguide with w=3.0 μm. The waveguide shown is approximately 2.6 mm in length. Since the intensity of the scattered light is weak and hardly changes with propagation distance, we can expect that the propagation loss is relatively small. Figure 5 shows the measured near-field profile from a strip-loaded waveguide with w=3.0 μm operated at 1.053 μm and the calculated intensity profile of the ${\mathrm{E}}_{11}^{\mathrm{y}}$ mode. We observed the intensity profile of the guided mode at the output using the IR camera with a 40X microscope objective. For comparison, we calculated the corresponding field distribution of the ${\mathrm{E}}_{11}^{\mathrm{y}}$ mode using a finite-element method [15]. The numerical calculation shows that the mode is well confined in the guiding layer under our experimental conditions. As seen in Fig. 5(a), the single-mode operation was confirmed.

 

Fig. 4. IR image of a characteristic streak of the scattered light from a strip-loaded waveguide with w=3.0 μm.

Download Full Size | PDF

 

Fig. 5. (a) Observed near-field profile from a strip-loaded waveguide with d=1.2, w=3.0 and h=0.1 μm and (b) calculated intensity profile of the ${\mathrm{E}}_{11}^{\mathrm{y}}$ mode using a finite-element method.

Download Full Size | PDF

Next, we evaluated the propagation loss in the fabricated strip-loaded waveguides using the scattering technique. The intensity of the scattered out of the waveguides was measured as a function of distance along the axis of propagation to determine the propagation loss. The measurement was carried out with an optical fiber probe with a 200-μm core diameter. Figure 6 shows the dependence of the relative scattered intensity on propagation distance for two values of the strip width w. It should be noted that the difference in the scattered intensity between two waveguides has no meaning. We can see an intense scattering in the first 4 mm of the waveguide for two strip widths, which is due mainly to failure of light to couple into the waveguide mode, i.e. the ${\mathrm{E}}_{11}^{\mathrm{y}}$ mode. An efficient coupling is performed by matching the incident beam profile to the waveguide mode profile. In the case of Fig. 6, the contribution of such leaky modes for w=3.0 μm is larger than that for w=6.0 μm. This reason is that the spot size of incident beam on the end-face was closer to ∼6 μm than ∼3 μm or that there existed any defects on the end-face of the waveguide with w=3.0 μm. The solid line is the least-mean-squares fit to linearly decreasing data, the slope of which gives the power loss coefficient. The propagation loss is 0.5 dB/cm for w=3.0 μm and 6.0 μm. Considering that the modal field is almost confined in the guiding layer as shown in Fig. 5 (b), we can easily understand that the measured loss does not strongly depend on the strip width. The obtained propagation loss is also smaller than the absorption loss in the corresponding coefficient of bulk glasses. It seems that the measurement using Eq. (1) overestimates the absorption coefficient of the bulk glasses due to small thickness of the samples used here.

 

Fig. 6. Dependence of the relative scattered intensity (in dB) on distance along the waveguide for two values of the strip width w. The straight lines represent best fits to the experimental data. Note that the difference in the scattered intensity between two waveguides has no meaning.

Download Full Size | PDF

3. Conclusions

We have fabricated single-mode Ag-As₂Se₃ strip-loaded waveguides by the photodoping and lift-off method and have evaluated the propagation loss in the waveguides at a wavelength of 1.053 μm by the scattering technique. The addition of Ag into a guiding layer made of As₂Se₃ glass increases its third-order nonlinearity and eliminates or weakens a harmful photo-oxidation reaction of As₂Se₃ films. We have also measured the absorption coefficient of the bulk glasses for different Ag contents to gain guidelines for the propagation loss in the waveguides to be fabricated. The measured propagation loss in the single-mode waveguides with the Ag content x=3 at.% is 0.5 dB/cm at 1.053 μm. The measured loss is attributable to the absorption loss since the scattering loss in the strip-loaded waveguide is negligible for a high field confinement into the guiding layer under the loading strip. Therefore, we can expect a lower propagation loss in these waveguides at the optical telecommunication wavelengths of 1.3 and 1.55 μm. At these telecommunication wavelengths, we can also use Ag-As₂Se₃ glasses with higher Ag content (x=15-30 at.%) as waveguide materials to utilize higher nonlinearity. We are trying to develop high-speed all-optical devices using such an Ag-As₂Se₃ strip-loaded waveguide.