3 Integrated optics on planar substrate

  The concept of integrated optics was born in the 70's with the development of optical communications by guided waves. A major problem of transmission by optical fibers was the signal attenuations due to propagation and the need for repeaters to reformat and amplify the optical signals after long distances. The solution offered by classical optics was unsatisfactory and Miller (1969) suggested an integrated, all-optical component on a single chip, with optical waveguides to connect them.

3.1 Principle of guided optics

  

Figure 3: Principle of optical guidance (see text for details)

For sake of simplicity, we first consider the wave propagation of a collimated incident beam into a planar waveguide. This particular structure is formed of three step-index infinite planar layers (see Fig. 3). Light can be observed at the structure output provided that total reflection occurs at each interface and constructive interferences occur between two successive reflected wavefronts (A and C in the figure). The first condition implies that a high-index layer is sandwiched between two low-index layers and gives the range of acceptable incident angle. The second condition translates into a phase difference between the wavefronts A and C multiple of 2$\pi$. Therefore the range of acceptable incident angles is no longer continuous but discrete. A single-mode waveguide is a guide which can propagate only the direction parallel to the waveguide. The core layer thickness ranges between $\lambda/2$ and $10\lambda$ depending on the index difference. Multimode guide propagates beams coming from different directions.

In practice, one needs the full electromagnetic field theory to compute the beam propagation inside the waveguide. The continuity relations of the electromagnetic fields at each interface lead to the equations of propagation of guided modes [Jeunhomme (1990)]. Depending on the wavelength and the guide thickness (l in the Fig. 3), these equations have either no solution (structure under the cutoff frequency), either only one solution (single-mode structure) or several ones (multi-mode structure). The number of solutions also depends on the difference of refractive index between the various layers of the structure. The larger the index difference are, the better the modes are confined. These equations also allow to estimate the energy distribution profile which can be approximated, to first order, by a Gaussian function. The major part of the energy lies in the channel, but evanescent waves can interact with evanescent waves coming from other close waveguides (see the directional coupler in Sect. 3.3).

In interferometry, multi-mode guided structures cannot be used since there exist optical path differences between the various modes. In the following, only single-mode waveguides are considered.

3.2 Current technologies

3.2.1 Ion exchange

  

Figure 4: Waveguide manufacture by ion exchange technique [Schanen-Duport et al. (1996)]

A first method to build integrated guides on planar substrate is based on glass ion exchange [Ramaswamy & Srivastava (1988),Ross (1989)]: the $\mbox{Na}^+$ ions of a glass substrate are exchanged by diffusion process with ions ($\mbox{K}^+$,$\mbox{Tl}^+$ or $\mbox{Ag}^+$) of molten salts. The local modification of the glass chemical composition increases the refractive index at the glass surface. A three-layer structure (air/ions/glass) is created and the light is vertically confined. By standard photo-masking techniques (see Fig. 4), the ion exchange can be limited to a compact area and create a channel waveguide. Since ion exchange only occurs at the surface of the glass, the last step of the process consists in embedding the guide, either by forcing the ions to migrate with an electric field or by depositing a silica layer on the waveguide. We obtain a component which guides the light like an optical fiber, the ion exchange area being the core and the glass substrate being the cladding. According to the ions of the molten salt, the refractive index difference can vary between 0.009 and 0.1 (see Table 1). This technology provides various components for telecom and metrology applications.

  

Table 1: Ions characteristics in ion exchange technology

3.2.2 Etching technologies

Another method consists of etching layers of silicon of various indices [Mottier (1996)]. These layers can be either phosphorus-doped silica or silicon-nitride. Both technologies can create channels by etching layers of material, where light is confined like in an optical fiber (see Fig. 5). The channel geometry is defined by standard photo-masking techniques. According to the fabrication process, $\Delta n$ can be either high (0.5) for very small sensors, or very low (between 0.003 and 0.015) for a high coupling efficiency with optical fibers. These technologies usually provide components for various industrial applications (gyroscopes, Fabry-Pérot cavities or interferometric displacement sensors).

  

Figure 5: Waveguide manufacture by etching technique [Mottier (1996)]

3.2.3 Polymers

Single mode waveguides made by direct UV light inscription onto polymers are in progress. Such a technology is still in development and the components present usually high propagation losses [Strohhöfer et al. (1998)].

3.3 Available functions with integrated optics

 

  

Figure 6: Available elementary integrated optics functions

The first two technologies provide many standard functions for wavelengths ranging between 0.5 $\ensuremath{\mu\mbox{m}}$ and 1.5 $\ensuremath{\mu\mbox{m}}$ (standard telecom bands). Several examples are presented (see Fig. 6):

1.

The straight waveguide is the simplest component.

2.

The curved waveguide allows some flexibility to reduce the size of integrated optics components. Its characteristics depend on the radius of curvature.

3.

The direct Y-junction acts as an achromatic 50/50 power divider.

4.

The reverse Y-junction is an elementary beam combiner similar to a beam-splitter whose only one output is accessible.

5.

The mirror is an Y-junction coupled with curved waveguides creating a loop. A straight transition between the Y-junction and the loop ensures a symmetrical distribution. The modes propagating through the loop in opposite directions interfere and then light goes back in the input straight waveguide.

6.

The directional coupler consists in two close waveguides. According to their proximity and the length of the interaction area, modes can be transfered between them and a power divider can be realized. The power ratio obviously depends on the distance between the two guides, the length of the interaction area and the wavelength.

7.

The characteristics of the X-crossing depend on the intersection angle. For high angles (e.g. larger than 10 degrees), the two waveguides do not interact: the crosstalk is negligible. For smaller angles, a part of power is exchanged between the two arms of the components.

8.

The taper is a smooth transition section between a single-mode straight waveguide and a multi-mode one. It allows light to propagate in the fundamental mode of the multimode output waveguide. The output beam is thus collimated.