1. Introduction

Conventional single-mode optical waveguides have a square-shaped core with a dimension of several micrometers to achieve circular optical field distribution that matches with a single-mode optical fiber. The optical mode field is strongly confined inside the core and decays exponentially outside (evanescent wave) [1]. In recent years, light propagation in subwavelength fiber attracts great attentions [2, 3]. Light can be largely trapped inside a sub-micrometer diameter silica fibers, but when the fiber diameter becomes smaller, large amount of light is distributed outside of the fiber and the effective mode field diameters is very large compared with the fiber diameter. For example, theoretical calculation showed that 90% of the optical power of a 216 nm-diameter silica fiber is distributed outside at 633nm wavelength [4]. In silicon-on-insulator waveguide study, it is also found that when a sub-micrometer Si waveguide is tapered to a nanometer tip, its mode field keeps the single-mode distribution but is very delocalized from the tip core to match well with a commercial single-mode fiber (mode field diameter is about 5 µm) [5]. In both cases, the sizes of waveguides are very small. Generally, when the size of a waveguide decreases to be close to the cutoff condition (0 for a symmetric waveguide structure), large part of the optical field extends to the claddings and is thus weakly confined. These waveguides are named weakly confined waveguides. The achievements in weakly confined waveguides definitely open a new route for compact integrated optical circuit fabrication. High intensity light in the cladding also leads to new applications such as ultra-sensitive optical sensors, ultra-short couplers, etc. [6, 7].

In this paper, a novel weakly confined waveguide structure called semi-weakly confined waveguide is reported. The waveguide has elongated rectangular core with a very thin (sub-micrometer) thickness. The optical mode field extends deeply into the buffer/cladding layers in vertical direction but is confined by the width of the core laterally. This waveguide can support a nearly circular waveguide mode and match well with single-mode optical fibers. Comparing with conventional waveguides, this novel waveguide geometry has two obvious advantages. One is that it is much easier to be fabricated, as very thin and wide cores can be easily grown and patterned. The other is the possibility of fabricating devices utilizing evanescent waves in vertical direction, like optical couplers, sensors and optical trapping platform etc. [6–8]. This novel weak-confined waveguide structure is capable of being a new building block in integrated optics.

2. Device design and fabrication

 

Fig. 1. The schematic structure of the optimized semi-weakly confined EDWA. The dashed and solid lines are the effective optical mode profile at 1550 nm and 980 nm respectively. Thickness of waveguide core and double buffer/cladding layers are labeled as well with the buffer/cladding that surrounds the core contains Er/Yb ions.

Download Full Size | PDF

To demonstrate the basic properties of this novel waveguide structure, an erbium-doped waveguide amplifier (EDWA) was designed and fabricated. Figure 1 schematically shows the EDWA structure. A thin and wide core is surrounded by buffer/cladding doped with erbium ions. Therefore, the property of this kind of EDWA depends sensitively on its optical mode confinement: the more the optical field extends to claddings, the higher the optical gain will be. Optical mode profile was calculated by FD-BPM method (finite difference beam propagation method) through BeamPROP software (RSoft Design Group, Inc.). By maximizing the optical power in the buffer/cladding and minimizing the coupling loss between the EDWA and a 980/1550 nm wavelength division multiplexer (WDM), the optimized waveguide core size of 5 µm wide ×350 nm high was found. The refractive index difference of the core and the buffer/cladding is Δn=0.05 (n_cladding=1.453) at 1550 nm. The calculated coupling losses between the EDWA and a 980/1550 nm WDM are 0.07 dB at 1550 nm and 0.63 dB at 980 nm, respectively. In Fig. 1, the calculated effective mode field profile is plotted as well. The optical power distributed in the buffer/cladding is 89.9% at 1550 nm and 72.9% at 980 nm. As the effective optical mode width in vertical direction at 1550 nm (2.69 µm) is larger than that at 980 nm (1.25 µm), the signal gain is achieved only in an effective gain area where there is population inversion of erbium ion generated by the pump light. Presumably this area should be a little less than the effective optical mode area of the pump light. High loss of the signal light occurs outside the effective gain area where rare earth ions are not pumped efficiently. To avoid absorption of signal light outside the effective gain area, a double-layered buffer/cladding structure was used in experiment. In the new structure, a 2.09 µm buffer/cladding layer that surrounds the core is doped with Er/Yb ions, while the rest of buffer/cladding is a passive one. About 85% of the 980 nm laser energy is inside the 2.09 µm active buffer/cladding layer as calculated by using the FD-BPM. This structure will help the EDWA to achieve better active properties.

 

Fig. 2. Processes to fabricate the EDWA sample. All layers were deposited by sol-gel method and DC-RTA technique.

Download Full Size | PDF

The processes to fabricate the EDWA samples are demonstrated in Fig. 2. A layer-by-layer DC-RTA (dip-coating and rapid thermal annealing) technique was used, in which each sol-gel layer of about 50 nm thick was dip-coated and rapidly annealed [9]. Two buffer layers, a 5-µm-thick passive layer and an 870-nm-thick active layer were deposited firstly. The sol for passive layer was prepared from tetraethyl othosilicate (TEOS), isopropyl alcohol, diluted HCl acid, 5 mol% P₂O₅ and 5 mol% Al(NO₃)₃. The sol for active layer was prepared by dissolving 0.5 mol% Er(NO₃)₃ and 0.5 mol% Yb(NO₃)₃ into the passive sol. Detailed process is similar to ref. [9]. The refractive index difference of the two layers is very small (~0.001). Yb³⁺ is added here as a photo-sensitizer for the pump light. However, as the total amount of Er³⁺ and Yb³⁺ in the matrix is quite limited, higher Yb³⁺ obviously reduces the total amount of Er³⁺, which in return will reduce the device gain. Thus, we used Er:Yb=1:1, this ratio is also used in literature [10]. Si wafers with 5-µm-thick oxidized silica layer were used as substrates. The core layer was dip-coated from a sol with a composition of 79.0 mol% methacryloxypropyl trimethoxysilane (MAPTMS), 10.5 mol% zirconium n-propoxide (ZPO) and 10.5 mol% methacrylic acid (MAA) [11]. The waveguides with different widths (3.5~8.5 µm) were patterned by standard photolithography and were wet etched. Two cladding layers, a 1.22-µm-thick active layer and a 5-µm-thick passive layer, were deposited on the core as shown in Fig. 1. The samples were annealed at 1000 °C for 0.5 h at the end.

3. Results and discussion

The coupling and propagating losses of the EDWA were measured by cutback method. The gain property of the EDWA was measured by a standard bi-propagating pumping scheme [12] with pump power up to 200 mW (100mW+100mW). In the measurement, a signal laser from a tunable laser source (1480~1630 nm, ANDO AQ4321D) was attenuated to -30 dBm and coupled into the EDWA through a 980/1550 nm WDM. The amplified signal was sent to an OSA (optical spectrum analyzer, ANDO AQ6317B). Gains of EDWA with different waveguide lengths were measured.

 

Fig. 3. Plots of insertion loss versus waveguide length at 1555 nm and 974 nm respectively.

Download Full Size | PDF

The EDWA sample with a width of 4.5 µm has the minimum measured coupling loss to a WDM. Figure 3 shows the measured insertion loss changing at 1555 nm and 974 nm with different waveguide lengths. The slope of the fitted curve gives the propagation loss, while the intercept is the total coupling loss. The measured coupling loss is 0.40 dB/facet at 1555 nm and 0.37 dB/facet at 974 nm, which is very close to the theoretically optimized value (5 µm width). It proves that the mode profile of the EDWA matches with that of the WDM quite well.

The propagation loss of the EDWA at 1555 nm is 1.10 dB/cm. It includes both the absorption and other losses such as scattering. Figure 4 shows the wavelength dependent insertion loss result for a waveguide of 3.5 cm long. Note that at the wavelength of 1580 nm, absorption of Er³⁺ ions can be ignored, thus from the insertion loss at that wavelength (2.83 dB) and the coupling loss obtained from Fig. 3 (0.8 dB), the extra loss except absorption is estimated to be 0.55 dB/cm. A cleaner fabrication environment and a better etching technique (like reactive ion etching, RIE) will be helpful to further decrease the loss.

 

Fig. 4. Gain spectra of a 3.5-cm-long semi-weakly confined EDWA with the double-layered buffer/cladding structure. The lines from bottom to top are gain spectra under total pump power from 0 mW to 200 mW respectively.

Download Full Size | PDF

 

Fig. 5. Plot of signal gain versus waveguide length at the peak near 1535 nm.

Download Full Size | PDF

 

Fig. 6. Plot of signal gain versus total launched pump power at the peak near 1535 nm. The waveguide length is 3.5 cm.

Download Full Size | PDF

Figure 5 plots the signal gain at the peak near 1535 nm versus waveguide length. The best gain result was obtained from an EDWA with a waveguide of 3.5 cm long. The gain spectra of this EDWA under different launched pump power are shown in Fig. 4. A maximum internal gain of 1.9 dB at 1535 nm was obtained with 200mW (100 mW +100 mW) pump power. Figure 6 is the relation between the signal gain at the peak near 1535 nm with the total launched pump power. Note that in contrast to conventional EDWA, the waveguide in this work does not have gain material in its core and the optical gain is totally realized from evanescent wave pumping. Further improvement of amplifier efficiency is possible by reducing loss and experimentally optimizing the double-layered buffer/cladding structure.

To demonstrate the necessity of the double-layer buffer/cladding design, a 3.7-cm-long EDWA with a uniform distribution of Er/Yb ions in the whole buffer/cladding was also fabricated. This EDWA has similar core dimension as the previous one. The gain spectrum at a total 200 mW pump power was measured and is shown in Fig. 7. In this case, however, the Er absorption valley is not inverted, as the wing of 1550 nm signal light profile is absorbed.

 

Fig. 7. Gain spectrum of a 3.7-cm-long semi-weakly confined EDWA with uniform distribution of Er/Yb ions in whole buffer/cladding and pumped at a total power of 200 mW.

Download Full Size | PDF

4. Summary

In summary, we showed that the semi-weakly confined waveguide structure is capable of extending large part of propagating light in one direction and achieves small coupling loss with a standard single-mode fiber. A double-layered buffer/cladding EDWA with evanescent wave amplification was designed and prepared by sol-gel method. Small coupling loss of 0.4 dB/facet and a maximum internal gain of 1.9 dB was measured. This new waveguide structure will be useful as optical sensors and building blocks for 3-dimentional integration.