CN1771446A - A Practical Approach to Beam Shaping and Reducing Losses Induced by Connecting External Light Sources and Optics to Thin Silicon Waveguides - Google Patents

Abstract

This paper describes a practical approach to achieving and maintaining high-efficiency transmission of light from input and output free-space optics into sub-micron-thick, high-refractive-index waveguides. The required optical components and methods for fabricating, calibrating, and assembling these components are discussed. Reliably maintaining high coupling efficiency within a realistic range of device operating parameters is discussed in the context of preferred embodiments.

Description

A Practical Approach to Beam Shaping and Reducing Losses Induced by Connecting External Light Sources and Optics to Thin Silicon Waveguides

Cross-reference citations to related applications

This application claims the benefit of Provisional Application No. 60/461,697, filed April 10,2003.

technical field

This invention relates to coupling devices associated with thin silicon optical waveguides, and more particularly to methods of beam shaping and reducing losses caused by connecting external light sources and optics to the thin waveguides.

Background technique

In the application of many devices, the input signal must be pre-processed in the device in order to optimize the device-specific technology for implementing the basic function; similarly, the signal from the core of the device must be post-processed before being transmitted to the outside. in order to produce a signal compatible with typical user requirements. For optoelectronic components, the required optical signal processing includes functions such as light generation, wavelength control, polarization control, phase control, beam direction control, beam shaping, beam splitting or recombining, modulation, and detection. For ease of use, or to control parameters that are critical to device performance, many pre-processing or post-processing functions can generally be integrated into the component. An important benefit, for example, is that the optical insertion loss of a device can often be reduced by integrating more optical functions into the component. This is not only because the choice of components can be more easily optimized for device know-how, but also because the physical connections between different devices or components are reduced. A low-loss optoelectronic component can be used in system applications because it is easier to apply in different places of the system and expand the application range of the system. Furthermore, the physical size of the device can be reduced through device integration.

The integration of pre- and post-processing optical functions is particularly critical for silicon-based optoelectronic circuits operating at infrared wavelengths. Because silicon lasers are not yet widely used in commercial applications, it is not yet possible to incorporate the light source in the same silicon as the signal processing and receiving components. Therefore, an optical signal must be introduced into the silicon wafer from an external light source. This requires the insertion of optical elements (between the light source and the waveguide) to pre-process the signal so that light of comparable intensity can be transmitted to the waveguide. Furthermore, because silicon-based detectors suitable for infrared wavelengths are just beginning to be developed, the optical signal must be transmitted from the silicon waveguide to an external detector or receiving element. Therefore, optical components are required at the output of the device to post-process the optical signal. Typical prior art methods for coupling light into high index contrast waveguides include prism couplers, grating couplers, wedge mode converters, and specially shaped fiber terminations or lensed fibers. While all of these optics have been used in a laboratory setting to transmit a portion of light from an external light source into a high-index-contrast waveguide, there are still challenges when these parts are used in prototypes or final products of low-loss devices. Lots of restrictions.

For example, specially shaped fiber terminations, lensed fibers, or wedge-mode converters can produce minimum beam spot sizes around 1.5 μm, which is not compatible with some submicron-sized silicon waveguides. In particular, single-mode silicon waveguides with dimensions on the order of 0.35 μm or less are required in many applications. Mismatches between the mode field diameter of the output beam of a tailor-made fiber or wedge mode converter and the mode field diameter in the waveguide mode will cause high insertion loss. Even if the diameter of the waveguide is on the order of a few micrometers, since the input and output ports of the device must lie on the cleaved facets of the wafer die containing the waveguide, the geometry of the device (e.g., the device layout and size) have many limitations.

The above limitations can be addressed by coupling light from an external source into or out of a high index contrast waveguide using a grating coupler or a prism coupler. With proper design, light can be successfully coupled into waveguides with thicknesses ranging from tens of nanometers to tens of micrometers. In addition, grating or prism elements may be positioned at any suitable location on the surface of the mold or wafer such that light can enter a substantial portion of the mold or wafer.

Although they have considerable advantages, difficulties in the manufacture of grating and prism couplers still limit their use in some special applications. The coupling efficiency of grating couplers is sensitive to grating period, depth and tilt angle. Theoretically, if the design goals of the grating parameters can be met, a coupling efficiency of about 70-80% can be obtained; in practice, due to the sensitivity to manufacturing tolerances, the measured coupling efficiency is mostly around 40%.

In the prior art, prism couplers required the placement of a large bulk optical element (a few millimeters in size) very close to the waveguide and precisely positioned relative to the waveguide. Here, "very close" means that the spacing between the optical element and the waveguide allows evanescent wave coupling of light from the optical element to the waveguide. For infrared wavelengths used in telecommunications applications, typical spacing values fall in the range of 200-500nm. The motion control required to manipulate the prism relative to the waveguide (e.g., using piezoelectric mounts) can be accomplished in a laboratory optical bench or test setup, but this approach cannot be realized in small optoelectronic components. Therefore, prism coupling applications are mainly limited to waveguide testing and qualification.

Because prism couplers have not been used in small optoelectronic components in the prior art, optical and mechanical components suitable for use with prism couplers in small device configurations have not been developed. For example, the prior art does not disclose specific embodiments of prism coupler arrangements for transmitting light into compact optoelectronic components or typical optical elements for receiving light therefrom. In laboratory setups, when the signal introduced into the prism coupler undergoes certain changes (such as changes in wavelength, polarization state, beam position, angle of incidence, etc.), the optical components can generally be adjusted in various ways to optimize signal transmission. For small devices, it is appropriate to design a device that is transparent to multiple inputs; that is, when the input state of a signal changes, only a small number of parameters need to be adjusted (or none at all) for the device to work properly. Therefore, the choice of optical parameters related to input and output beams, input and output optical elements, and prism couplers directly affects the versatility and manufacturability of the device. However, because prism couplers have not been integrated into small optoelectronic devices in the prior art, detailed design schemes for making versatile and manufacturable devices have not been developed.

Therefore, it is also necessary to design and implement an optical system that can interface with a prism coupler in a small, low-loss, and stable optoelectronic component in the art.

Contents of the invention

An unsolved need in the prior art addressed by the present invention involves the design of an optical system for processing infrared light signals entering and exiting a compact prism-coupled optoelectronic device.

In particular, the present invention specifies several embodiments of optical elements that provide the necessary interface for permanently coupled miniature prism and waveguide components. These interfaces include, but are not limited to: free-space optics that direct light from an external source into a high-index prism structure, optics or structures fabricated on the same silicon wafer or mold with etched sides serving as the input and output sides of the prism , an evanescent wave coupling layer that forms the direct physical interface between the high-index prism and the waveguide, and a free-space optical element that receives the output beam exiting the output facet of the prism.

The various embodiments described above are particularly suitable for thin silicon waveguides in the wavelength range commonly used in telecommunication applications. However, the various interface devices of the present invention are equally applicable to other devices, and may use larger sized waveguides and/or other wavelength ranges. Specific embodiments of the launch optics and conditions that provide a new and compact packaging solution for prism coupling devices are specified. Designs are disclosed that minimize the end-to-end insertion loss of small optoelectronic devices using prism coupling, and theoretical coupling efficiencies for specific embodiments are calculated. More advantageously, a specific and manufacturable embodiment of an evanescent wave coupling layer that produces a desired output beam intensity profile and reduces insertion loss is specified.

The advantages of reducing the required free-space beam size relative to manufacturing requirements will become apparent in the course of the ensuing description, with reference to the accompanying drawings.

Description of drawings

Now please refer to the attached picture.

Figure 1 shows a silicon-based prism coupler permanently affixed to a silicon-on-insulator (SOI) wafer containing a silicon waveguide layer.

Figure 2 shows the geometric route of a beam of light propagating within a prism structure, including the emission angles inside and outside the prism (corresponding to θ _air and θ _Si , respectively), and the physical dimensions of the optical coupling region at the prism surface, the The prism surface is directly connected to the evanescent wave coupling layer;

Fig. 3 shows the external beam emission angle θ _Si of the prism of the embodiment in Fig. 1 in a certain telecommunication wavelength range, and the range under the condition of three different silicon waveguide thicknesses;

Fig. 4 has shown the range of the beam launch angle θ _air (outside the prism) in the air of the embodiment in Fig. 1 within a certain telecommunication wavelength range and under the condition of three different silicon waveguide thicknesses;

Figure 5 shows the full range of emission angles inside the prism structure (θ _Si ) and outside the prism face (θ _air ), covering the device silicon thickness range from 0.1 to 0.21 μm, and the wavelength range from 1290 to 1590 nm;

下的角半高宽(用FWHM(θ_air)表示)；Figure 6 shows the coupling efficiency curves over a certain free-space input beam diameter value as well as at three different coupling constant values

The angular half-maximum width under (expressed in FWHM(θ _air ));

Figure 7 shows the coupling efficiency of the embodiment in Figure 1 as a function of silicon oxide evanescent coupling layer thickness, simulation results for three different waveguide layer thicknesses in a silicon-on-insulator wafer;

Fig. 8 shows the coupling efficiency as a function of the thickness of the evanescent wave coupling layer for an embodiment similar to that in Fig. 1, the simulation results for three different materials making up the evanescent wave coupling layer.

FIG. 9 shows the maximum shift in flatness (termed "wedge angle") as a function of the free-space input beam diameter, which is consistent with the theoretical model of constant thickness for the evanescent coupling layer of the embodiment in FIG. 1 .

Figure 10 shows the optimum wedge angle as a function of free-space input beam diameter for an embodiment similar to that of Figure 1 with a wedge-shaped evanescent wave coupling layer.

上的输入光束尺寸与输入自由空间光束尺寸

的比值，在SOI晶片中四种不同波导层厚度下的情况；Fig. 11 shows that for the embodiment in Fig. 1, the prism coupling surface in a certain telecommunication wavelength range

The input beam size on and the input free-space beam size

The ratio of , in the case of four different waveguide layer thicknesses in the SOI wafer;

Figure 12(a) and (b) show in top and side projections how an initially unpolarized input beam is converted into two separate beams with desired polarization directions, thus allowing light to pass through a The prism structure efficiently couples into the waveguide;

Figure 13 shows an example of using actuated MEM micromirrors to control the proper launch angle of a light beam from a horizontal light source to the outside of the prism.

Figure 14 shows a physical layout that demonstrates side emission of light entering the device from an edge emitting diode or other fiber exiting from one fiber, with the output on the opposite side of the assembly;

Figure 15 shows a physical layout that demonstrates side emission of light from an edge-emitting diode or other fiber input entering the device from one fiber, with the output on the same side of the assembly;

Figure 16 shows an example of using an arrayed VCSEL light source and a microprism array to direct the beam to a prism structure;

Figure 17 is another embodiment of the device of Figure 16, wherein an array of edge-emitting diodes is used to replace the array of VCSEL light sources;

Figure 18 is another embodiment similar to Figure 16 and using a set of lensed optical fibers co-located with beam steering means;

Figure 19 shows a prism wafer that includes additional optics to collimate and manipulate the beam prior to the evanescent wave coupling layer interface;

Figure 20(a) shows the preferred embodiment of Figure 1 with an evanescent wave coupling layer of constant thickness, and Figures 20(b) and (c) show the input and output beam amplitudes as a function of z, Fig. 20(d) shows a superposition of Figures 20(b) and (c)(c);

Figure 21 contains a schematic diagram showing a particular device for forming a semi-Gaussian wave; and

Figure 22(a) shows the preferred embodiment of Figure 1 with an evanescent wave coupling layer that varies linearly in thickness, and Figures 22(b) and (c) show the input and output beam amplitudes as a function of z, Figure 22(d) shows a superposition of Figure 22(b) and (c);

Detailed ways

To better understand the gist of the present invention, it is important to understand the requirements associated with an input beam that is first transmitted to the input facet of a typical prism structure 10 shown in Figure 1 and then coupled through the prism structure to a thin silicon waveguide 12. A detailed schematic of the propagation of the input beam in the prism structure is shown in Figure 2. The light beam enters the prism structure 10 through the hypotenuse (input face) surface 14, which is coated with an anti-reflective layer 16 to reduce the transmission from a low-refractive-index medium (atmosphere) to a high-refractive-index medium (in the embodiment in FIG. 1 ). Fresnel loss caused by the transmission of silicon). Referring to FIG. 2 , the input light beam makes an incident angle θ _air with the normal to the incident face surface 14 and is then refracted by the prism. For consistency with known optics, it is most convenient to express the angle (θ _Si ) inside the prism as the angle between the beam and the axis perpendicular to the waveguide. According to the geometric relationship in Figure 2, θ _Si and θ _air have the following relationship:

θ _Si = θ _pr -sin ^-1 {sinθ _air /n _Si },

Wherein, for the wavelength n _Si in the 1.3-1.6 μm band, the refractive index of silicon ≈3.5.

对于耦合效率，此光束在棱镜和消散波耦合层界面上的投影是关键的参量。从与图2相关的几何关系中可见，输入自由空间光束和直径为 的自由空间输入光束在棱镜耦合表面上的投影的关系可表示为：Refraction simultaneously expands the size of the beam inside the prism, along the axis shown in Figure 2, by a factor of:

For the coupling efficiency, the projection of this beam on the interface of the prism and the evanescent coupling layer is a critical parameter. From the geometric relations associated with Fig. 2, it can be seen that the input free-space beam and diameter are The relationship for the projection of the free-space input beam on the prism coupling surface can be expressed as:

Figure 2 illustrates the geometrical constraints governing the propagation of light beams inside and outside the prism 10, and Figure 1 shows a preferred layout in which the prism coupler is fabricated from a single silicon wafer and permanently affixed to a connecting link containing the waveguide 12. on the SOI wafer 20. As shown in FIG. 1 , the waveguide layer 12 is separated from the silicon substrate 22 by a blocking oxide layer 24 . The required prismatic surfaces are fabricated on the silicon wafer through a combined wiring and etching process, rather than using a discrete precision prismatic optical element. The desired portions of the vertical walls 30, 32 can be formed by a variety of etch processes, with the prism slopes 14, 18 most easily formed by an anisotropic wet etch process. Anisotropic processing has different etch rates for different crystal planes, so the prism slopes 14, 18 form specific angles to the wafer plane. For the structure in Figure 1, the silicon prism wafer has a <100> crystal orientation, so anisotropic KOH etching will result in a crystal plane at an angle of 54.74 degrees to the wafer plane. An evanescent wave coupling layer 26 is obtained by depositing a layer of material with a lower refractive index than silicon (n _Si ≈ 3.5) on the upper waveguide surface of the silicon prism wafer or connected SOI facet. The prism coupler is then permanently attached to the SOI wafer containing the waveguide, preferably using a semiconductor bonding process, although adhesive and solder bonding methods can also be used. In the resulting prism coupler/SOI wafer part, the base of the prism coupler 10 (prism coupling surface 15) is in direct contact with the waveguide surface 12 of the SOI wafer 20, thus resulting in a prism/evanescent wave coupling layer/waveguide sandwich. To reduce Fresnel losses at the sloped surfaces of the input and output prisms (hereinafter referred to as "prism faces"), an additional layer (or layers) of material is deposited on the surface of the silicon prism coupler with integrated prism faces. This layer or multilayer structure acts as an anti-reflective coating 16 which significantly increases the transmission of light as it traverses the prism face.

By using theories known in the prior art, the beam angle θ _Si in a silicon prism structure can be calculated over a range of waveguide thicknesses compatible with single-mode propagation and wavelength bands used for telecommunication applications. The calculated results of _θSi for waveguide thicknesses of 0.1, 0.14 and 0.21 μm and wavelength range of 1290–1630 nm are shown in Fig. 3. These typical waveguide thicknesses were chosen because optical and high-speed electronic functions can be integrated in these relatively thin waveguides. It can be seen that the beam angle θ _Si (defined in Figure 2) covers the range of 38° to 58° over the desired wavelength and waveguide thickness range. To determine a suitable launch angle θ _air for the exterior of the prism, the previously obtained relationship of θ _Si and θ _air can be used. As before, for the embodiment in FIG. 1 , for a <100> oriented silicon wafer, using an anisotropic etch process that produces input and output angled facets would result in θ _pr =54.74 degrees. However, the use of the embodiment in Fig. 1 is not limited to this particular value of _pr ; any other value of _pr obtainable by an etching process or a different method may also be used. Figure 4 shows the calculated results of _θair for the wavelength range 1290-1630nm and waveguide thicknesses of 0.10, 0.14 and 0.21μm. The range of incidence angles in air is much larger, varying from -15 degrees to 90 degrees; this is due to the large difference in refractive index between air (n ≈ 1.0) and silicon (n ≈ 3.5).

Figure 5 provides a graphical representation of the range of angles inside the prism (θ _Si ) and outside the prism (θ _air ) that must be achieved for a prism of 54.74 degrees in order to allow the use of the device to cover a wide range of wavelengths and waveguide thicknesses. full range. The air emission condition can be realized over a wide range of wavelengths and waveguide thicknesses, except that wavelengths greater than 1590 nm cannot be used for a waveguide thickness of 0.10 μm. Thus, the main advantages of the embodiment shown in FIG. 1 include (1) coordinating common semiconductor wiring, etching and bonding processes to produce manufacturable prismatic couplers and waveguide devices, and (2) forming a useful structure, This structure can be used for applications covering a wide range of infrared wavelengths and waveguide thicknesses.

The usefulness of the device illustrated in FIG. 1 can be further enhanced by selecting optical and spatial properties of the input and input beams that simplify connection to the optical signal interface of the device in FIG. 1 . Because the wavelength range and power of the input signal are usually determined by the application, the polarization direction, beam shape, beam (or wavefront) quality, and propagation direction can be adjusted within the component. For lens coupling applications, according to the present invention, precise control of these parameters is necessary in order to achieve the desired high coupling efficiency of light from the lens coupler into the waveguide. In particular, the following conditions must be met:

(1) The input beam must be determined by the polarization state and wavelength of the input beam, the refractive index and thickness of the silicon device waveguide layer 12 (hereinafter represented by W) and the evanescent wave coupling layer 26, and the refractive index of the prism 10 and its surrounding medium Angle of incidence exits. If the incident beam emerges from a suitable angle of incidence, the wavefield propagation constants in the prism 10 and waveguide 12 will match, so that high coupling efficiency can be obtained.

(2) The beam must be highly collimated at the prism coupling surface 15 so that the finest part of the input Gaussian beam falls near the prism coupling surface. It is known that the coupling efficiency is reduced if the phase of the wave front varies greatly over the projection range of the wave on the prism coupling surface 15 .

的位置上，以使耦合效率最大化。光束被垂直侧壁34截断的一小部分被完全内反射，在从输出面18出射前，首先由垂直侧壁34反射，然后由棱镜耦合表面15反射。需要强调的是，相对此位置 的小偏移会使耦合效率稍稍降低(约10％)。用这种特殊的方式截断输入光束在棱镜耦合表面15上的投影，防止了从棱镜结构传输至波导12的光耦合回到棱镜结构。(3) The input beam must intersect the prism coupling surface 15 at a specific position, which depends on the form of the evanescent wave coupling layer and the beam intensity distribution of the input optical signal. For a Gaussian input beam and an evanescent wave coupling layer 26 of constant thickness, it can be seen that the center of projection of the beam on the prism coupling surface must be located at a distance from the vertical side wall 34 of the prism shown in FIG. 2

position to maximize coupling efficiency. The fraction of the beam intercepted by the vertical side wall 34 is fully internally reflected, first reflected by the vertical side wall 34 and then by the prism coupling surface 15 before exiting the output face 18 . It should be emphasized that relative to this position A small offset of 0 decreases the coupling efficiency slightly (about 10%). Truncating the projection of the input beam onto the prism coupling surface 15 in this particular way prevents the light transmitted from the prism structure to the waveguide 12 from being coupled back into the prism structure.

从先有技术中已经知道，通过实现输入光束的投影(约为

和主要由消散波层厚度决定的耦合强度参数(此后称为“α”)间的一种特定关系，耦合效率可达到最大。这是因为α和 是重叠积分中的重要参数，该重叠积分决定了耦合效率。(4) In order to maximize the coupling efficiency, the thickness of the evanescent wave layer must be suitable for the size of the projection of the input beam on the coupling surface of the prism,

It is known from the prior art that by realizing the projection of the input beam (approximately

The coupling efficiency can be maximized by a specific relationship between the coupling strength parameter (hereinafter referred to as "α") mainly determined by the thickness of the evanescent wave layer. This is because α and is an important parameter in the overlap integral that determines the coupling efficiency.

To meet these conditions in small optoelectronic components, suitable collimating, shaping, and beam redirecting microelements, as well as additional polarization and phase control optics, are important for the coupling efficiency of light into the structure in Figure 1. Because the typical dimensions of the prism facets in Figure 1 are about 0.5-1.0 mm, the optics' diaphragms must be of similar size to keep the overall assembly compact. The maximum dimension of the beam must be slightly smaller than the size of the optics to prevent transmission losses due to beam confinement. As discussed below, other manufacturing factors specific to prism coupling applications impose more stringent constraints on the maximum beam size. For efficient prism coupling, there is an optimal beam size (related to the properties of the evanescent coupling layer, as described earlier) and a minimum beam size such that the beam remains collimated as it traverses the prism structure and intersects the prism coupling surface .

Manufacturing tolerances for the device shown in Figure 1 are more easily met if an appropriate maximum beam size is chosen. In particular, significant advantages are obtained with respect to the tolerance of the launch angle of the input light beam I and the thickness of the evanescent wave coupling layer 26 .

时，可得到80％的最佳耦合效率。α是表示耦合强度的参数，也是从棱镜结构的输出面出射的光束的形状的特征常数，单位为长度的逆，出射光束的形状的形式为g(z)∝exp(-αz)。参数α主要由消散波耦合层厚度、消散波耦合层中的传播常数以及波导两个边界处的反射引起的相位变化决定。It is known from the prior art that for an evanescent wave coupling layer of constant thickness, when

, the best coupling efficiency of 80% can be obtained. α is a parameter representing the coupling strength, and is also a characteristic constant of the shape of the beam emitted from the output surface of the prism structure. The unit is the inverse of the length, and the shape of the emitted beam is in the form of g(z)∝exp(-αz). The parameter α is mainly determined by the thickness of the evanescent wave coupling layer, the propagation constant in the evanescent wave coupling layer, and the phase change caused by the reflection at the two boundaries of the waveguide.

值减小，α必须增大，相当于更强的耦合或更薄的消散波耦合层。增强的耦合强度引起更宽的谐振，而更宽的谐振允许更宽的波长范围或等效地输入角度耦合进入波导。实际上，β空间中(β表示传播常数)谐振的洛伦兹分布图的半高宽(FWHM)直接正比于α，关系为：If set to 0.68 to optimize coupling, then because of

As the value decreases, α must increase, corresponding to stronger coupling or thinner coupling layers for evanescent waves. The enhanced coupling strength induces wider resonances which allow a wider range of wavelengths or equivalently input angles to be coupled into the waveguide. In fact, the half-maximum width (FWHM) of the Lorentz distribution diagram of resonance in β space (β represents the propagation constant) is directly proportional to α, and the relationship is:

并依照以下关系：FWHM(β)=FWHM(n _Si sinθ _Si )=αλ/π. The numerator and denominator are multiplied by

And according to the following relationship:

θ _Si ＝θ _pr -sin ^-1 {sinθ _air /n _Si }, it can be obtained that the FWHM as a function of the input angle θ _air is:

其中，

in,

F(θ _air ，θ _pr ){1-(sinθ _air /n _Si ) ² } ^1/2 /[cos(θ _air )×cos{θ _pr -sin ^-1 (sinθ _air /n _Si )}].

For a particular device configuration, such as that shown in Fig. 1, θ _pr and W (waveguide thickness) are fixed quantities (θ _pr ) ₀ and W ₀ , respectively. In addition, if a specific wavelength λ ₀ is selected, the central value of θ _air is also set at a specific value (θ _air )0 (as shown in FIG. 4 ). In this case, the intensity width at half maximum for small deviations from θ _air (external launch angle into the prism structure) can be expressed as:

线性增加，并随着光束直径在棱镜耦合表面15上的投影的倒数增加。对于一个给定的耦合效率 的值，一定输入角范围内的强度分布可通过减小输入光束的直径在棱镜耦合表面15上的投影来增强。同样，稍提高耦合常数

的值，一定输入角范围内的强度分布可得到提高，而耦合效率仅稍稍下降。从制造方面考虑，选择合适的耦合常数 和光束投影 十分重要，以便最终的装置在其使用期内对于小的θ_air变化更加稳定。下面的一个例子示出了适合于高耦合效率的光束尺寸和输入角的变化范围。This means that the intensity distribution within a certain range of input angles varies with the parameter that determines the coupling efficiency

increases linearly and with the inverse of the projection of the beam diameter onto the prism coupling surface 15 . For a given coupling efficiency The value of , the intensity distribution within a certain range of input angles can be enhanced by reducing the projection of the diameter of the input beam on the prism coupling surface 15 . Likewise, slightly increasing the coupling constant

For a value of , the intensity distribution within a certain range of input angles can be improved with only a slight decrease in the coupling efficiency. From manufacturing considerations, choose the appropriate coupling constant and beam projection is important so that the final device is more stable to small changes in θ _air over its lifetime. An example below shows a range of beam sizes and input angles suitable for high coupling efficiencies.

值和三个不同耦合效率

值的函数。这四个选择的光束尺寸对应于以下情况：(1)63μm：透镜型光纤部件的标准输出光束尺寸；(2)100μm：激光组件内集成了微透镜的垂直共振腔表面发射激光器(VCSEL)的典型光束尺寸；(3)200μm：标准光纤光学准直器(与一个GRIN或非球面镜对齐的光纤/套管部件)中可获得的最小光束尺寸；以及(4)360μm：标准光纤光学准直器(与一个GRIN或非球面镜对齐的光纤/套管部件)中最常使用的光束尺寸。为了计算输入发射角变化下的耦合效率和半高宽，光束在棱镜耦合表面上的投影

用前面的公式从自由空间光束直径

计算出来。接下来考虑通过调整

的值来改变耦合效率的影响。如果消散波耦合层比给定光束尺寸的最佳值厚，系统将处于欠耦合状态，即

小于最佳值。对于 对于图1中的实施例，仍可获得72％的耦合效率。对于输入角的容差，这种情况不是十分适合，原因是谐振更为锐利，且θ_air变化的容差小于最佳耦合条件下的容差。对于图1中工作在1550nm波长下的装置，欠耦合条件下72％的耦合效率对应于约40nm的过厚消散波耦合层(见图7)。可以看出，此耦合值下对于任何可实现的构造，FWHM(θ_air)一般不超过0.35度。最佳耦合条件下，时FWHM(θ_air)已增加至0.4-0.6度，而对于更大的光束直径则保持在约0.1-0.2度。现在考虑消散波耦合层约40nm过薄的情况，则出现耦合效率为72％， 的过耦合情况。从图6可以看出，对于

角度容差已显著提高至0.7-1.1度，对于更大的光束直径则达到约0.2-0.35度。因此，自由空间光学器件校准后，适度过耦合的装置中使用小的光束直径可显著降低装置对工作时或装置老化时产生的小的变化的敏感程度。Figure 6 shows the FWHM(θ _air ) as four free-space beam diameters

value and three different coupling efficiencies

function of value. The four selected beam sizes correspond to the following: (1) 63 μm: standard output beam size for lens-type fiber optic components; (2) 100 μm: vertical cavity surface emitting laser (VCSEL) with microlenses integrated in the laser assembly Typical beam size; (3) 200 μm: smallest beam size achievable in standard fiber optic collimators (fiber/tube components aligned with a GRIN or aspheric mirror); and (4) 360 μm: standard fiber optic collimators (fiber/ferrule assembly aligned with a GRIN or aspheric mirror) is the most commonly used beam size. To calculate the coupling efficiency and full width at half maximum for varying input launch angles, the projection of the beam onto the coupling surface of the prism

From the free-space beam diameter using the previous formula

Calculated. Next consider adjusting the

value to change the effect of the coupling efficiency. If the evanescent wave coupling layer is thicker than optimal for a given beam size, the system will be undercoupled, i.e.

less than the optimum value. for For the embodiment in Fig. 1, a coupling efficiency of 72% is still obtained. For the tolerance of the input angle, this situation is not very suitable, because the resonance is sharper, and the tolerance of the change of θ _air is less than the tolerance of the optimal coupling condition. For the device in Fig. 1 operating at a wavelength of 1550 nm, the coupling efficiency of 72% under the under-coupling condition corresponds to an excessively thick evanescent coupling layer of about 40 nm (see Fig. 7). It can be seen that FWHM(θ _air ) generally does not exceed 0.35 degrees for any achievable configuration at this coupling value. Under optimal coupling conditions, The FWHM(θ _air ) has been increased to 0.4-0.6 degrees for larger beam diameters while remaining at about 0.1-0.2 degrees for larger beam diameters. Considering now that the evanescent wave coupling layer is too thin at about 40nm, the coupling efficiency is 72%. overcoupling. From Figure 6, it can be seen that for

The angular tolerance has been significantly improved to 0.7-1.1 degrees and to about 0.2-0.35 degrees for larger beam diameters. Thus, the use of small beam diameters in moderately overcoupled devices after calibration of free-space optics can significantly reduce the sensitivity of the device to small changes in operation or as the device ages.

的值偏离最佳值0.68。作为一个例子，图7中显示了图1中的优选实施例的耦合效率在波导层12的三种不同的厚度下作为氧化硅消散波耦合层26的厚度的函数。消散波耦合层的厚度参考1550nm的应用波长和直径为63μm的输入自由空间光束作估计。图中显示的装置层厚的范围代表了目前绝缘体上硅处理中的层厚的实际分布范围。目标装置层厚为0.14μm，如优选实施例中所示。由图可见，消散波耦合层的厚度必须落在目标值±20nm的范围内，此例中约为320nm，以防止耦合效率降低10％(如果考虑波导层12的厚度的容差，±0.01μm)。尽管如此，±20nm的容差必须在光束在棱镜耦合表面上的投影的整个物理范围内得以保持，以确保高的耦合效率。这项条件将更容易得到满足，如果(1)选择构成消散波耦合层的介质，使得图7中的耦合效率曲线的宽度适当；(2)将棱镜耦合器固定到SOI晶片的波导表面上的处理方法使得厚度容差在光束投影的物理范围内得以保持；以及(3)光束在棱镜耦合表面上的投影的物理范围相对很小。An additional benefit of using a relatively small beam diameter arises from the limited physical extent to which the beam interacts with the evanescent wave coupling layer. In order to obtain high coupling efficiency, the thickness of the evanescent wave coupling layer must be precisely controlled. Changes in layer thickness will directly translate into changes in α, making

The value of deviates from the optimal value of 0.68. As an example, the coupling efficiency of the preferred embodiment of FIG. 1 is shown in FIG. 7 as a function of the thickness of the silicon oxide evanescent wave coupling layer 26 at three different thicknesses of the waveguide layer 12 . The thickness of the evanescent wave coupling layer is estimated with reference to an applied wavelength of 1550 nm and an input free-space beam of diameter 63 μm. The range of device layer thicknesses shown in the figure represents the actual distribution of layer thicknesses in silicon-on-insulator processing today. The target device layer thickness is 0.14 μm, as shown in the preferred embodiment. As can be seen from the figure, the thickness of the evanescent wave coupling layer must fall within the range of the target value ±20nm, approximately 320nm in this example, to prevent the coupling efficiency from reducing by 10% (if considering the tolerance of the thickness of the waveguide layer 12, ±0.01 μm ). Nonetheless, a tolerance of ±20 nm must be maintained over the entire physical range of the beam's projection on the prism coupling surface to ensure high coupling efficiency. This condition will be more easily satisfied if (1) the medium constituting the evanescent wave coupling layer is selected such that the width of the coupling efficiency curve in Fig. 7 is appropriate; The processing method allows thickness tolerances to be maintained within the physical extent of the projection of the beam; and (3) the physical extent of the projection of the beam onto the prism coupling surface is relatively small.

Figure 8 shows the results of an analysis similar to Figure 1, but showing the coupling efficiency as a function of the evanescent coupling layer thickness for three different evanescent coupling layer refractive indices. These three values represent three different typical dielectrics: air (n≈1.0), silicon oxide (n≈1.45), and silicon nitride (n≈2.0). The basic form of the coupling efficiency curves in these three cases is the same, but it is obvious that the thickness of the optimal evanescent wave coupling layer varies, and the width of the coupling efficiency curve slightly broadens with the increase of the refractive index of the evanescent wave coupling layer. Referring to Figure 8, when n=2.0, the thickness of the evanescent wave coupling layer must fall within the range of the target value ±20nm, about 385nm in this example, to prevent the coupling efficiency from decreasing by 10% (if considering the silicon waveguide layer (refer to Figure 1 The thickness tolerance of the waveguide layer marked), ±0.01μm). Therefore, there is a small benefit to be gained by using a higher index evanescent wave coupling layer. Interestingly, all three media (air, silicon dioxide, and silicon nitride) work well within the scope of the current embodiment as long as the correct thickness of the evanescent wave coupling layer is obtained. The indicated coupling curve widths (±20nm for silicon oxide and ±25nm for silicon nitride) correspond to a tolerance of ±6-7% for the thickness of the evanescent wave coupling layer, which matches current manufacturing methods.

对于图1中的装置构造，对应的光束在棱镜表面的投影约为610μm。相似的计算给出，允许的楔形角已降低至0.04μm/610μm＝6.6×10^-5弧度，或0.004度。大部分楔形角容差的改进源于以下事实，即对于较小的光束尺寸关键的间隙间距需要在较小的范围内保持。因为上述所有允许的楔形角都非常小，且随着光束在棱镜耦合表面上的投影反向减小，图1中显示的可高效率地耦合光的设备的制造由于使用了相对较小的光束尺寸的设计方案而显著改进了。For the device configuration shown in Figure 1, if a bundle diameter The 63μm free-space input beam is transmitted to the input prism surface, and the input beam is on the coupling surface The largest dimension of the projection on is about 110 μm (for a wavelength of 1550 nm, the thickness of the waveguide is 0.14 μm, and the thickness of the silicon oxide evanescent wave coupling layer is about 320 nm). In addition, it can be seen from FIG. 7 that the thickness of the evanescent wave coupling layer can be varied by ±20 nm, while still maintaining a coupling efficiency of more than 70% for the same device configuration. During device fabrication, the prism coupling surface is generally not absolutely parallel to the waveguide plane. A small offset from the parallel position will cause the magnitude of the evanescent wave coupling layer thickness to vary slightly along the prism coupling surface. Fig. 9 shows, for the embodiment shown in Fig. 1, a shift in relative parallel position that can be supported over a range of input beam sizes, yet is still consistent with a model of an evanescent wave coupling layer of equal thickness. As shown in FIG. 7, for the evanescent wave coupling layer as a coupling region with a substantially constant thickness, a maximum ±20nm, or total 40nm, thickness variation in the optical coupling region can be supported. Therefore, if the projection of the input beam on the prism coupling surface is 110 μm, the maximum allowable wedge angle is about 0.04 μm/110 μm = 4×10 ⁻⁴ radians or 0.02 degrees. If a silicon nitride evanescent wave coupling layer is used, similar calculations show that for a free-space beam with a diameter of 62 μm, the maximum offset of the allowable flatness can be appropriately increased to 0.026 degrees. If a large beam size is used, the optimal thickness of the evanescent wave coupling layer will increase, but the variation in the thickness allowing high coupling efficiency remains essentially the same, about ±20nm. For a free-space beam size of 360 µm

For the device configuration in Fig. 1, the projection of the corresponding beam on the prism surface is about 610 μm. Similar calculations give that the allowable wedge angle has been reduced to 0.04 μm/610 μm = 6.6×10 ⁻⁵ radians, or 0.004 degrees. Much of the improvement in wedge angle tolerance stems from the fact that the critical gap spacing for smaller beam sizes needs to be maintained over a smaller range. Because all of the above-mentioned allowable wedge angles are very small and decrease inversely with the projection of the beam on the prism coupling surface, the fabrication of the device shown in Figure 1 that couples light efficiently due to the use of relatively small beam The size of the design solution has been significantly improved.

In one variation of the device configuration shown in Figure 1, a small variation in the thickness of the evanescent wave coupling layer along the input and output light coupling regions can improve the coupling efficiency by more than 80%. It is known from the prior art that the gradient thickness of the evanescent wave coupling layer is such that the intensity of the output free-space beam is Gaussian in nature, and the thickness of the optical coupling region where the beam is first separated from the waveguide by the output prism is greater than the optimal thickness , while the thickness of the light-coupling region at the point where the last remaining light intensity in the waveguide is separated by the output prism is less than optimal. This is different from the case of an evanescent wave coupling layer of constant thickness, where the intensity distribution of the output beam is exponential. The improved mode matching between the Gaussian input beam and the essentially Gaussian output beam brought about by the wedge-shaped evanescent wave coupling layer increases the theoretical coupling efficiency from 80% to about 97%. Without a detailed mathematical discussion, basic information for calculating the appropriate angles for the wedges in Figure 10 can be obtained from Figure 7 . As mentioned earlier, the parameter α, which is related to the coupling strength and appears in the functional form of the intensity distribution of the output beam, is mainly determined by the thickness of the evanescent wave coupling layer. For a wedge-shaped evanescent wave coupling layer whose thickness (z) varies along the propagation direction of light in the waveguide, the coupling strength at a given value of z is directly related to the local value of α, α(z). The thickness of the evanescent wave coupling layer must change from weak coupling (small α(z)) to strong coupling (large α(z)) over a distance scale approximately equal to the projection of the input beam on the prism coupling surface. To obtain high coupling efficiency, it is necessary to determine a suitable mean thickness value (this yields a value of α close to the optimal value of α, which is suitable for optimal coupling of a constant thickness evanescent wave coupling layer) and a suitable value of thickness with z Linear variation, or "wedge angle". For the example of FIG. 7, when W=0.14 μm, it can be seen that the coupling efficiency drops to about 37% (or 1/e) of its maximum value at evanescent wave thicknesses of 250 nm and 450 nm. This corresponds to a total 200 nm change across the projection of the beam on the coupling surface. For the embodiment in Fig. 1 and the configuration shown in Fig. 7, when the total projected beam length is 110 μm, the electric field amplitude of the projection of the output beam on the coupling surface varies from the distance peak down to about 37% of its maximum value. It can be deduced from general theory and detailed overlap integrals in the prior art that this matching will give rise to a high degree of overlap between the output and input beams. Therefore, the optimal gradient of the linearly changing gap is 200nm/100μm=1.8×10 ^-3 rad=0.1°. Note that this condition also holds for the relatively small (but still achievable) free-space diameter of 63 μm. If a silicon nitride (Fig. 8, n ≈ 2.0) is used instead of a silicon oxide evanescent wave coupling layer, the optimum slope increases modestly to 0.13° for a free-space beam diameter of 63 μm.

Figure 10 shows the exact same calculations for the beam size used in the calculations shown in Figure 9, resulting in the optimal wedge angle as a function of the free-space beam size. Note that the optimal wedge angle increases by a factor of 6–7, from 0.02° at a free-space beam size of 360 μm to 0.10° at a free-space beam size of 63 μm. Again, the tolerance improvement is primarily due to the need to maintain accurate thickness variations over a smaller distance for smaller beam sizes.

Because the wedge angle required for both constant thickness and slope thickness evanescent wave coupling layers is relatively small, increasing the wedge angle significantly improves the manufacturability of the final device. It can be seen from Figures 9 and 10 that the tolerance of the required wedge angle begins to improve as the free-space beam diameter decreases below 200 μm. Additional benefits can be gained as the free-space beam size decreases below 100 µm.

定义，其中n为光束传播穿过的介质的折射率，而其他符号与前面的定义相同。物理上，瑞利路程大致对应于光束保持准直的距离。当使用棱镜结构将光从外部光源传输到波导时，为了获得高耦合效率，光束腰必须位于输入光束在棱镜耦合表面的投影的附近。在与棱镜耦合表面相交之前，光束必须在硅棱镜耦合器内(通常在空气或其他输入光学器件内)传播一定距离。如果光束尺寸太小，空气、输入光学器件和硅内允许的路径长度将太小，以致实际中无法实现。接下来将依照图1中的设备构造的上下文详细说明一个示例计算。While the preceding discussion points out that reducing the beam size can significantly improve the manufacturability of the device for a number of reasons, the size and layout of the prism coupler (and any input optics before it) limits the minimum beam size. Small diameter beams diverge rapidly over relatively small propagation distances. A commonly used figure of merit for this distance is called the Rayleigh path (denoted here as z _R ), and is given by the relation

Definition, where n is the refractive index of the medium through which the beam propagates, and the other symbols are the same as the previous definitions. Physically, the Rayleigh path roughly corresponds to the distance over which a beam remains collimated. When using a prism structure to transmit light from an external source to a waveguide, for high coupling efficiency, the beam waist must be located near the projection of the input beam on the prism coupling surface. The beam must travel a certain distance within the silicon prism coupler (usually within air or other input optics) before intersecting the prism coupling surface. If the beam size is too small, the allowed path lengths in air, input optics, and silicon will be too small to be practical. An example calculation will next be detailed in the context of the device configuration in FIG. 1 .

If the base size of the prism structure shown in Figure 1 is 0.45mm (measured horizontally from the deepest part of the V-shaped groove to the corner edge produced by the etching process), and the light with a wavelength of 1550nm is transmitted from the silicon prism coupler to For a θ _Si value of 45.5° to launch, the corner beam must travel a path length of approximately 400 μm from the input prism face to the prism structure and prism coupling surface. The launch distance prior to the prism face must be included in the calculation of the position of the beam waist. This launch distance includes the path length of the beam in air, as well as the thickness of the optics used to precondition the beam. Depending on the number of elements required, the beam path leading to the input facet ranges from 1 mm (reasonable manufacturing tolerance for device alignment) to a few mm. Because the refractive index of air (and more generally, the refractive index of the input optics) is much lower than that of silicon, and the path length preceding the prism facet usually exceeds the path length in the prism structure, the Rayleigh path calculation is mainly performed by prior Depends on the emission from the input prism face. Using the _zR relationship given above, it follows that for a beam diameter of 20 μm, the Rayleigh path is 0.2 mm in air and 0.7 mm in silicon. For the larger beam diameter of 63μm, the Rayleigh path is about 2.1mm in air and 7.3mm in silicon. For a beam diameter of 100 μm, the Rayleigh path is about 5.1 mm in air and 17.6 mm in silicon. In order to obtain a transmission distance of the order of several millimeters in air, calculations have shown that it is feasible to use beams with a size of the order of 60-100 μm.

Because typical micro-optical elements such as polarizing beam splitters, waveplates, and micro-wedges or prisms can be 0.5 mm thick or less, there are elements that can be used to shape, steer, and adjust the polarization of the beam after a collimating lens direction. Therefore, a beam size of 60-100 μm meets the requirements for small beams, miniaturized components, and input columns for micro-optics. To simplify packaging and other assembly aspects discussed earlier, it is appropriate to choose a design range of input beam diameters on the order of 60-100 μm.

定义为自由空间中的光束直径)。典型情况下，光束在棱镜耦合表面上的投影

超过自由空间光束直径一至三倍。A final consideration regarding the input beam size is that the input beam size is the lower bound on the beam size imposed by the prism coupler, evanescent wave coupling layer, and waveguide. The properties of these three components determine the angle of the beam inside the prism, θ _Si , and thus directly affect the projection of the beam on the coupling surface of the prism, (See Figure 2). In addition, the material and geometry of the prism coupler will follow the relation Determines how the beam is refracted on the input prism face. It should be pointed out that, in general, due to the refraction at the slope and the projection at the surface of the coupling surface,

defined as the beam diameter in free space). Typically, the projection of the beam on the prism coupling surface

One to three times more than the free-space beam diameter.

Figure 11 shows the amplification of the beam on the coupling surface of the prism due to these effects for different device layer thicknesses in Figure 1 and for the full telecommunications wavelength range. It can be seen that in most cases the beam is magnified by a factor of 1.6-2.0 along the propagation axis. Larger values and their more rapid increases (eg W = 1.0 μm for θ _pr =54.74°) correspond to increased refraction effects at high oblique angles of incidence on the prism slope. For the same reason, these structures are not very suitable from an assembly point of view. Therefore, for practical considerations, it will be assumed that the beam size at the coupling surface of the prism is increased by a factor of 1.4-2.4 relative to the free-space value. Also, note that by choosing a specific waveguide thickness and prism angle (eg W = 1.7 μm, θ _pr = 54.74°), the projection of the input beam on the coupling surface can be substantially wavelength independent. In this way, a suitable small beam size can be obtained for any wavelength in the wavelength range, thus simplifying the design of the device. This allows a given prism wafer/evanescent coupling layer/waveguide configuration to be used with high coupling efficiency over a much wider range of wavelengths than an arbitrary device configuration.

It should be understood that many different components can be designed and installed to generate, transmit and condition the optical signal in order to obtain the desired beam characteristics in the present invention. The following description includes typical configurations of light sources and columns of optical elements that may provide a convenient interface to a device similar to that of FIG. 1 .

Furthermore, for some applications, the properties of the prism coupler, evanescent wave coupling layer, and waveguide can be chosen to simplify interfacing with external light sources or receiving elements. In particular, some components used to transmit and condition the input optical signal can be designed inside the prism coupler wafer or chip, thus reducing the total number of independent components and simplifying the assembly process. By choosing the materials, thicknesses and geometries of the evanescent wave coupling layer and the waveguide, favorable emission geometry and beam shape can be obtained, which also simplifies the assembly process.

Laser diodes are typical light sources commonly used in optoelectronic devices using telecommunications wavelengths (1.1-1.65 μm). Many infrared laser diodes typically consist of a multilayer structure of gallium arsenide-based or indium phosphide-based materials, with light emitted from the cleaved edge facet of the laser chip (technically known as an edge-emitting laser diode). The laser diode can be used directly in the form of the chip, or, as in various prior art packaging techniques, the laser chip can be connected to an output fiber through a series of optical elements. A second class of typical laser diodes is known in the art as vertical-cavity surface-emitting lasers, or VCSELs. Infrared VCSELs consist of a multilayer structure (using gallium arsenide, indium phosphide or indium gallium arsenide-based materials) where light is emitted perpendicular to the layer stack and through the top surface of the device.

For some applications, free-space optics need to be used to transfer from the laser chip into the prism structure. Direct coupling to the laser enables very compact packaging and provides a high degree of polarization control. However, due to the small emitting surface of the laser and the long infrared wavelength, the output beam may diverge severely. Edge-emitting laser diodes operating in the 1300-1600nm range typically have about 32°-50° FWHM beam divergence in the direction perpendicular to the junction and about 10°-25° FWHM beam divergence in the direction parallel to the junction Diverge.

Because the beam divergence is large and anisotropic, at least two lenses are required to achieve efficient free-space beam collimation with high wavefront quality. In one lens assembly, a pair of crossed cylindrical lenses is used to correct for astigmatism and to provide collimation in the fast and slow axes. To effectively collimate the highly divergent or "fast" axis, the first cylindrical lens is usually made of a graded index material (technically called a "GRIN" lens). A second cylindrical lens that collimates the less divergent or "slow" axis can be made from a variety of optically clear materials, since lens shaping alone is not sufficient to provide collimation. The diameter of the input beam emerging from a typical laser diode followed by a miniature GRIN rod lens can be chosen to fall within the range of 40 μm to a few mm. In the second configuration, the first lens is used to reduce the divergence angle perpendicular to the junction direction until its value is equal to the divergence angle parallel to the junction direction, rounding the beam and correcting astigmatism. Such lenses are sometimes called "laser diode correctors" or "circularizers". The second lens can now be a conventional collimating microlens (to reduce beam divergence to near zero), such as a plano-concave or aspheric mirror, and can be made of a variety of optically transparent materials. The advantage of the second configuration is that only one special lens is required instead of two. The output beam diameter from a typical laser diode followed by a correction lens can be chosen to fall within the range of 100µm to 1mm.

VCSELs emit moderately divergent beams with divergence angles ranging from 29° (for lens-type components) to 18°. The lenses used can be conventional collimating microlenses (to reduce beam divergence to near zero), such as plano-concave or aspheric mirrors, and can be made of a variety of optically transparent materials. To obtain collimated beams with small beam diameters, integrated microprisms can be introduced as part of the VCSEL structure itself. Thus, for a VCSEL active area of 3 μm, a collimated beam with a diameter of 100-200 μm can be obtained. Although VCSELs at mid-infrared wavelengths (1270-1650 μm) are just beginning to emerge, they have potential advantages in reducing the component count and complexity of optoelectronic components.

In other applications, a length of fiber optic can be used as a conduit for passing light from a laser source to a prism coupler. If the laser source is located in a separate housing with a fiber output, the prism-coupled waveguide must be provided with an input fiber component that can be directly connected to the output of the laser source. (If there are multiple fiber optic devices between the laser source and the prism-coupled waveguide, the input fiber optic component of the prism-coupled waveguide must be connected to the terminal fiber output of the link). If the laser source is introduced into the same assembly as the prism-coupled waveguide, for some applications it is still advantageous to use interleaved fibers between the laser chip and the prism coupler. For example, a wider range of collimated beam sizes and shapes can be achieved using a specially terminated fiber. This special termination can be applied on the end of the fiber closest to the prism coupler, and usually involves shaping one end of the fiber, or fusing a microprism directly to the end of the fiber. By varying the size and radius of the bend at the end of the fiber or the lens, a collimated beam of minimal spot size (also called "beam waist") can be obtained at a user-specific working distance. Using current process technology, fiber collimators with beam waist diameters in the range of 15 μm to 100 μm can be fabricated in this way. The laser light source can be connected to the other end of the optical fiber using a lens assembly as described in detail in the prior art. Thus, for the configurations of Figures 13, 14 and 15, a 60 diameter beam can be produced from the fused lens/fiber assembly for both fiber and laser inputs.

While the lens assembly provides the necessary beam collimation, it is still necessary to ensure that the beam is in the proper polarization state before entering the prism. Although both transverse electric (TE) and transverse magnetic (TM) polarization states can be efficiently coupled into the waveguide, at a particular value of θ _Si , only one polarization state can be efficiently coupled. Since an edge-emitting laser diode emits a beam with a known polarization state that is stable, a microwave plate can be used to rotate the polarization state to the appropriate state. For some applications, the waveplate can be omitted entirely by choosing an appropriate polarization state consistent with emission from an edge-emitting diode. If the input beam is passed through a polarization maintaining fiber, the fiber can be rotated during assembly to ensure the proper polarization state, again eliminating the need for additional polarization optics.

However, the polarization state of a VCSEL is not known exactly. In particular, the polarization state may not change with time, but the direction is unknown, or conversely, the polarization state may change with time or laser drive current. Similarly, if a non-polarization-maintaining input fiber is used, the polarization state of the light will be indeterminate and drift with time. An element used in a prior art optical circulator can also be used in the present invention to obtain the correct polarization state, as shown in FIG. 12 . The input beam is passed to a birefringent element 50 which splits a single input beam into two polarized beams: one in the desired polarization state and the other perpendicular to the desired polarization state. Because the refractive index of the light is different for the two polarization states, the two light beams initially propagate within the device 50 in different directions. The beam in the desired polarization state continues to propagate in a medium that does not affect its polarization state. However, the beam whose polarization state is perpendicular to the desired polarization state passes through the second birefringent element, ie, the beam direction control element 52 which rotates its polarization state by 90 degrees to the desired polarization state. The final output is two separate beams, slightly offset from each other, both in the desired polarization state. In most applications, the two elements 50 and 52 are combined to form an optical subassembly that is easy to align and manufacture. Naturally birefringent materials such as _YvO4 , quartz, rutile or lithium niobate, or artificial birefringent elements such as subwavelength diffractive optics can be used. If the direction of the polarizing component is such that the two beams of light are irradiated on the prism surface 14 at exactly the same incident angle, both beams of light can be efficiently coupled into the waveguide layer 12 . For some applications, the light beams can be recombined after entering the waveguide layer. Recombination is easily accomplished by suitable guiding structures within the SOI waveguide layer 12 itself. Once the input beam is collimated and has achieved the desired polarization state, the optical signal must be launched from the prism face at the proper angle of incidence if the desired wavelength of light is to be efficiently coupled into the waveguide. For an embodiment as in Fig. 1, the beam can be launched directly into the prism structure at angle _Θair , or a small optical element can be used to redirect the input beam to an input angle _Θair on the prism structure. For packaging reasons, for edge-emitting diode sources, fiber-input, or vertical-cavity surface-emitting lasers (VCSELS), it is often convenient to launch the light parallel to the wafer (θ _air = -35.3° for direct launch). For an external light source, such as a VCSEL, it is also suitable for emission perpendicular to the waveguide (θ _air =54.74° for direct emission). It can be seen from Figure 4 that by choosing an appropriate waveguide thickness for a given wavelength, an appropriate launch condition can be selected. However, for some designs, the required waveguide thickness for a particular launch angle may not be compatible with competing device requirements. For these reasons it is appropriate to package some beam direction control optics close to the light source. In addition to angular selection, beam direction steering optics can be used with other alignment techniques, such as positioning of the light source relative to the prism, to ensure proper (translational) positioning of the beam on the prism.

Figures 13 and 15 detail typical methods of directing a beam from an edge-emitting diode or fiber optic to the face of a prism. In Figures 14 and 15, a collimated free-space beam from an edge-emitting diode or fiber is directed to a micro-optical prism or wedge. The magnitude of the beam deflection increases with the refractive index and wedge angle of the micro-optics. A similar micro-optics can be used at the output to direct the output beam to a receiving fiber. Alternatively, diffractive optical elements such as linear phase gratings can be used as beam direction steering elements. Diffractive optical elements are very effective in beam direction control applications because the dispersion effect of a well-designed grating can be large, allowing large polarization angles (up to 60°). Another advantage is that more complex diffractive optical elements can simultaneously perform more than one optical function, providing better performance with fewer elements. As an example, in addition to acting as beam direction steering elements, diffractive optical elements can be used for wavefront correction to improve wavefront quality.

In Fig. 13, a micromirror 54 fabricated by microelectromechanical systems (MEMS) processing is used to reflect light to a suitable angle of incidence θ _air . In the example shown in Figure 13, micro-hinges 56 fabricated using silicon micromachining methods hold the mirror in the correct angle and position. One benefit of using this technique is that the position and angle of the micromirror 54 can be manipulated and adjusted so that the θ _air and the position of the beam relative to the etched corner can be adjusted to maximize the light transmitted into the waveguide. As before, the same structure is used on the output side to direct the output beam to the receiving fiber.

Figures 14-19 illustrate specific input and output optical configurations that can be interfaced with prism-coupled waveguides with high coupling efficiency. Although a particular optical element (eg, lensed fiber 60 in FIG. 15) is shown in only one embodiment, it can be inferred that a given element may be conveniently used in a variety of different embodiments. Thus, the embodiments detailed in Figures 14-19 are examples only in nature and do not detail possible configurations.

Figures 14 and 15 show two conventional fiber optic pigtail configurations connected to prism-coupled waveguides. While the prism structure and SOI device wafer are joined to form one component, the packaging of the input and output optics columns may comprise a single component. In this case, the optics are housed and calibrated in holders on a separate carrier, which in turn is connected and calibrated to the prism/SOI device waveguide components. Alternatively, if the prism structure is fabricated on a silicon wafer, additional masking and etching processes can be used to define recesses that mount the free-space element in the surface of the silicon wafer opposite the connection surface. In both cases, grooves of a size roughly equal to the outer dimensions of the optical element are machined into the substrate material. The free-space optics are then positioned, aligned, and fixed in place in the trench. In Figures 14 and 15, the optical signal enters and exits the assembly through a single optical fiber (technically called "pigtailed fiber").

In Figure 14, there are two important separate devices in a pigtailed optical fiber assembly. The input fiber optic interface and the output fiber optic interface are on one side and the other side of the assembly, respectively. The embodiment shown in Figure 14 uses a polarization maintaining fiber 70, which ensures that the correct polarization state is obtained without other stray polarizations. A tiny optical lens (equivalent to a micro-sphere, a micro-GRIN lens, or a micro-spherical lens) is used to collimate the divergent light coming out of the optical fiber, and then the collimated light beam is directed to the beam direction control element 74, which The element deflects the light beam by an angle so that it is incident on the entrance face 76 of the prism 78 at the appropriate angle of incidence θ _air . If the beam control element 74 is further positioned on a separate submount and can be rotated as shown in FIG. 14, the incident angle can be adjusted during assembly and remains constant during the life of the device. At the output of the device, the output beam passes through the same sequence of optical elements in reverse order. Although polarization maintaining fiber is not strictly required at the output of the device, the use of polarization maintaining fiber allows the configuration in Figure 14 to be used as a bidirectional system.

The embodiment in Figure 15 is similar to that in Figure 14, showing an assembled device with the output port on the same side as the input port. This unique configuration is advantageous when the overall size of the assembly needs to be kept small. As shown in Figure 15, the direction of beam propagation is reversed by a reflective optical element located within the waveguide layer of the SOI wafer. The optical signal is introduced through an optical fiber 80 located at the bottom end of the module. In this configuration, a microlens 82 is fused directly to the fiber 80 so that a well collimated beam is obtained with a single subcomponent. Since the polarization state of the beam exiting the lensed fiber is unknown, the polarization control element 84 is used to convert the incident beam into two beams with the desired polarization state. The orientation of the polarization control element 84 is such that the two outgoing beams are horizontally displaced (ie the plane containing the two beams is parallel to the plane of the wafer). Because the distance between the two beams is small, on the order of hundreds of microns, the two beams can be deflected by the same beam direction control element 86 . The two beams of light are transmitted to the incident surface 88 of the prism 89 at exactly the same angle θ _air , and are coupled into the waveguide layer 12 of the SOI wafer. The two light beams, which are phase shifted from each other, are then recombined into a single beam by optical elements located in the waveguide layer. After passing through the remaining optoelectronic structures within the SOI waveguide layer, the exiting light emerges from the output prism face and propagates into a similar optical output column. However, unless an unpolarized output beam is desired again, the polarization control element at the output can be omitted.

Figure 16 shows an alternative implementation. Among them, a group of laser light sources 90 are directly integrated in one component. Because VCSELs emit light through the outer surface and can be small in size (about 100-250 μm), they can be easily aligned with a silicon wafer or a long strip of silicon prisms 94 etched in a wafer mold. As shown in Figure 16, beam collimation and beam direction control can be achieved by an array of refractive lenses 92 and diffractive lenses or beam steering elements. The size of the elements in the lens set can range from a few microns to a few millimeters. Most compact structures can be achieved by etching control prisms and/or collimating lenses directly into the VCSEL wafer itself. In this way, all light beams are transmitted to a prism set 94 at exactly the same angle θ _air , then coupled into the waveguide 12 , and exit from the output surface 95 of the prism bar 94 . A similar set of lenses and diffractive elements 96 are used to deflect, shape and converge the light beams onto an array of receiving ends of guide fibers 98 . Alternatively, edge-emitting diodes can be used in a similar configuration, as shown in Figure 17, as long as the spacing of the array is sufficient to accommodate slightly larger edge-emitting devices. Referring to FIG. 17, an embodiment utilizing an edge-emitting laser diode 91 further utilizes a laser diode collimating lens array 93 and is placed at the output of the edge-emitting laser diode array 91, where the collimating lens array is used for The beam direction control element 92 provides a suitable signal distribution. FIG. 18 shows another version of the embodiment of FIG. 17 with a lensed fiber array 97 placed at the output end of the collimating lens array 93 .

If it is desired to reduce the total number of components and adjustment steps, the required optical components can be processed in a silicon prism wafer or mold, as shown in Figure 19. In this configuration, the beam enters the silicon prism wafer 100 through any user-specific prism wafer 100 surface, rather than directing the incident beam at the appropriate angle θ _air to an etched "hypotenuse" prism input face. In the example of FIG. 19, the light beam enters through the surface 102 of the prism wafer 100, which is directly opposite the surface 104 to which the SOI wafer 106 is attached (the prism coupling surface). As the light beam propagates in the prism wafer 100, it encounters a series of surfaces that change its direction of propagation until the desired emission angle in silicon, θ _{Si ,} is achieved. These surfaces may consist of the top surface 102 and bottom surface 104 of the wafer 100 or surfaces formed by any other etching process. For a beam of light propagating in a silicon wafer, due to the high refractive index of silicon, total internal reflection can be obtained on these surfaces over a wide range of incident angles. For an air-silicon interface (assuming n≈1 for air and n≈3.5 for silicon), the required angle of incidence must be greater than the critical angle for total internal reflection of 16.6°, while for a silicon-silicon nitride interface (assuming nitrogen Refractive index of silicon dioxide n≈2), the required angle of incidence must be greater than 34.8°. If the incident angle is smaller than the critical angle, a high reflectivity can still be obtained by plating gold on a part 108 of the surface 102 as a reflector. Because the thickness of the silicon wafer is relatively small, about 500-700 μm, the light beam still encounters many different reflective surfaces while traveling a relatively short physical distance (about a few millimeters) within the silicon wafer 100 . Thus, the silicon prism wafer itself can be used as a compact low-loss beam direction steering element.

In the simplest configuration, a prism wafer is used to (1) direct the beam to a suitable angle _θSi , and (2) couple the beam into the waveguide. After entering through the top surface of the silicon prism coupler, the beam is refracted in the silicon wafer and incident on the etched surface. If the angle of incidence on the etched surface is large enough, then total internal reflection will occur on this surface. Conversely, a sufficiently small angle of incidence causes totally reflected light to be emitted towards the top surface. When the angle of incidence on the top surface is large enough, the beam will again undergo total internal reflection at the top surface. After total internal reflection at the top surface, the beam is launched towards the optical coupling region at a suitable launch angle θ _Si . This method of steering the beam is very efficient because it allows a wider launch angle θ _Si (due to the high refractive index of silicon) than direct launch from the top of the silicon prism coupler into the optical coupling region. Additional optical functionality can be added by incorporating optical elements on the top surface in the direct path of the light beam. In the example of Figure 19, the optical element can be located at the initial entry point of the beam into the top surface of the silicon prism coupler, or at the point of total internal reflection on the top surface. These optical elements may include, but are not limited to, the following: refractive or diffractive lenses to collimate diverging input beams, or other diffractive optical elements that provide additional beam direction control, beam shaping, wavefront correction, or polarization control capabilities.

The use of these refractive and diffractive elements provides additional optical functions such as collimation and polarization control to be integrated in silicon prism couplers. Microlenses can be fabricated in silicon using a combination of conventional lithography, photoresist vias, plasma etching, diffusion, and implantation techniques. Alternatively, grayscale lithography techniques can be used to produce more complex aspheric lens shapes. Many diffractive elements, ie grating structures, can be processed in silicon substrates using conventional lithographic techniques. However, higher resolution lithographic techniques such as electrolithography may be required to obtain subwavelength grating structures that can be used as polarization-controlling elements.

The coupling efficiency of the typical device in Figure 1 can be significantly improved by carefully considering how beam shape affects device performance. There are three main interfaces to consider here: (1) the shape of the free-space input beam from the input optics; (2) the exact form of the evanescent wave coupling layer; and (3) the shape of the free-space output beam as well as the output receive optics.

In general, coupling efficiency can be determined using overlap integrals well known from the prior art. From this integral it follows that 100% coupling efficiency can only be obtained when the input and output beam shapes match.

For the typical embodiment in Figure 1, three relevant overlap integrals need to be considered:

(1) η ₁ = Beam shape of the light source relative to the desired beam projection on the coupling surface of the prism

(2) η ₂ = beam shape on the coupling surface of the input prism relative to the beam transmitted from the coupling surface of the output prism

(3) η ₃ =shape of the beam transmitted from the coupling surface of the output prism relative to the desired beam shape of the output receiving optics.

Coupling efficiency is first discussed here in the context of the preferred embodiment shown in FIG. 20 . As can be seen from this embodiment, the input and output silicon prisms are separated from the silicon waveguide by an evanescent wave coupling layer of constant thickness and constant refractive index.

For the laser input and fiber output of a standard pigtailed fiber, the total coupling efficiency of the embodiment in Figure 20 is defined as:

η=η ₁ η ₂ η ₃ ≈64%

The coupling efficiency _η1 is determined by the loss caused by generating a well-collimated Gaussian beam from a light source such as a fiber or laser input. If the optics are integrated within the source (e.g. using a lensed fiber or laser source with integrated collimation and beam shaping devices), _η1 will be very high, close to 100%. The coupling efficiency, _η, is determined by the ratio of the power of the prism's free-space output beam to the power of the free-space input beam. However, for a Gaussian input beam in free space, _η2 will never exceed 80%. It is known from the prior art that _η2 is limited due to the different mode intensities of the input and output beams of this embodiment. The intensity of the input beam along the direction of propagation is Gaussian, while the intensity distribution of the beam from the output prism along the direction of propagation is exponential (see Fig. 20(b) and (c) for amplitude versus position plots). Finally, for the same reason, the coupling-to-output light efficiency _η3 is about 80%. Again, this is due to incomplete overlap of the exponentially enveloped free-space beam emerging from the prism with the desired Gaussian beam at the output of the fiber. thus,

η = η ₁ η ₂ η ₃ ≈ (1)*(0.8)*(0.8) = 0.64 or about 2dB of insertion loss.

Obviously, if the coupling efficiency of the embodiment shown in FIG. 20 is to be improved, further beam shaping is required to increase η ₂ or η ₃ to more than 80%. For a light source, such as a laser, the most common beam shape is a Gaussian or square wave distribution. It can be shown that both waveforms produce a coupling efficiency of η ₂ =80%. To increase η ₂ , it is obvious that the input beam must have an intensity distribution close to the exponential envelope of the beam exiting the output prism. One way to do this is to use a "semi-Gaussian" input waveform. As shown in FIG. 21 , an initial input Gaussian beam is incident on a wave splitter structure 120, and the center of the Gaussian beam is aligned with the intersection point of the wave splitter surface. In this way, the two half-beams are respectively transmitted to a prism (not shown) and coupled to a waveguide. One of the half-beams can be inverted using suitable optical elements such as a folding mirror 122 . It is important to ensure that the two half-beams do not recombine before entering the waveguide, which would result in interference fringes that strongly modulate the intensity distribution of the input beam. In this case, η ₂ =integral of overlap of the semi-Gaussian waveform and the output exponential waveform=97%. The coupling efficiency η ₁ can be reduced during conversion into two half-Gaussian beams. Clearly, if there is any significant advantage in adjusting the incident beam shape, there will be η ₁ >83%. Coupling efficiency η ₃ will be more difficult because the standard method used to generate a beam with a more Gaussian profile from the input beam would significantly reduce the intensity. For the structure shown in Fig. 20, it can be expected that if there is additional incident beam shaping, the maximum total coupling efficiency η can reach 80%; if there is no additional incident beam shaping, it can only reach about 64%.

A higher and easier to achieve overall coupling efficiency η is obtained in the embodiment shown in FIG. 22 . In this embodiment, the silicon prism is separated from the silicon waveguide by an evanescent wave coupling layer whose thickness varies linearly with position. On the input face, the thickness of the evanescent wave coupling layer when energy is first transmitted to the waveguiding layer is smaller than the thickness of the evanescent wave coupling layer when most of the energy has been transmitted to the waveguiding layer. On the output face, when most of the energy is still in the waveguide, the thickness value of the evanescent wave coupling layer is relatively large. Instead, the energy of the beam is reduced as it couples out of the waveguide layer and into the prism. Since the approximate Gaussian beam waveform can be maintained in the entire optical path from the incident light source to the exit fiber interface, this method can obtain higher coupling efficiency than the embodiment shown in FIG. 20 .

As mentioned earlier, a standard incident beam from a laser source or fiber is a collimated Gaussian beam with a high coupling efficiency _η . To increase the coupling efficiency η ₂ , the waveform of the free-space output light emitted from the prism must be closer to a Gaussian beam. Although the output light is usually not a true Gaussian beam, the coupling efficiency can still exceed 80% if the overlap integral of the new output light with the input Gaussian beam is larger than the overlap integral of the exponential envelope with the input Gaussian beam. It is known from the technical literature that one way to make the output beam more Gaussian is to gradually vary the thickness of the evanescent wave coupling layer with the direction of beam propagation. If the thickness of the evanescent wave coupling layer is constant, the beam will be coupled out of the waveguide into the prism with the same coupling strength at all points, and the result is that the output beam waveform can be written as g(z)∝exp(-αz) (see Figure 20 ). The coupling strength decreases with the increase of the thickness of the evanescent wave coupling layer, and increases with the decrease of the thickness of the evanescent wave coupling layer. If the output beam shape is closer to the input beam shape, the first light exiting the prism surface will be weakly coupled, so most of the light remains inside the waveguide layer. To ensure this, the evanescent wave coupling layer must be well above the optimum coupling value. To achieve this, the coupling strength of the light must be increased so that most of the light can be extracted, peaking the output "Gaussian" beam. Therefore, this part of the light must be quantified as the interface where the evanescent wave coupling layer is close to the optimal thickness. In this way, most of the energy is delivered from the waveguide and exits the system completely through the output prism. Although the coupling strength continues to increase as the thickness of the evanescent wave coupling layer decreases, the amount of light exiting the prism begins to decrease due to the decreasing energy of the light within the waveguide layer. In this way, an output light waveform more in line with Gaussian distribution can be obtained. Although the output beam is usually not a true Gaussian waveform, the overlap integral η ₂ of the new output beam with the input Gaussian beam is ≈97%. It is important that the slope of the evanescent wave coupling layer be of an appropriate value to produce the desired beam shape. The determination of this slope was discussed in the previous section on beam dimensions.

Because the free-space output beam from the prism has the Gaussian shape required for the output fiber interface, the coupling efficiency _η can now be very high. As mentioned earlier, the overlap integral η of the near-Gaussian beam emerging from the output prism with the Gaussian mode characteristic of the fiber _can be as high as 97%. If desired, collimating and rounding optics similar to those used to shape the laser diode beam prior to transmission to the fiber optic cable can be used to reduce any output beam divergence or ellipticity. Finally a lens is always required to focus the collimated beam onto the fiber. This lens may be an integral part of the lensed fiber or collimator assembly, or a separate spherical lens or gradient index lens may be used for this purpose with a common fiber termination.

The overall coupling efficiency of the embodiment shown in Figure 22 can be expressed as:

η = η ₁ η ₂ η ₃ ≈(1)*(0.97)*(0.97)≈0.94, or about 0.3 dB of loss. This is probably the easiest way to obtain efficient end-to-end coupling from a laser or fiber-based input to a fiber output, and this technique can be used in other applications that are more sensitive to insertion loss. However, improvements in coupling efficiency must be weighed against the additional demands of grayscale lithography required to obtain evanescent-wave coupling layers of varying thickness. It should be noted that any evanescent wave coupling layer structure that can generate a similar approximately Gaussian output beam or a more Gaussian output beam from the output prism can obtain a high coupling efficiency ≥ 94%. That is, the improvement in coupling efficiency is not limited to the evanescent wave coupling layer whose thickness varies linearly. For the purposes of this invention, "more Gaussian" may be defined as any output beam shape that increases the known overlap integral. For example, it can be shown that an evanescent wave coupling layer whose thickness varies logarithmically with distance along the waveguide produces a more Gaussian beam than one whose thickness varies linearly. (Plotting a curve of coupling efficiency versus layer thickness on a logarithmic coordinate will yield a more symmetrical coupling efficiency peak or curve). Fabrication of such thick structures is generally more complex, but is still necessary if the overall coupling efficiency of 94% is not sufficient for the application's insertion loss requirements.

Claims (46)

1. An optical coupling device providing a path for signals into and out of a silicon optical waveguide formed in a surface layer of a silicon-on-insulator (SOI) wafer comprising a silicon-on-insulator A silicon optical waveguide layer on an insulating layer on a substrate, the optical coupling device comprising A silicon-based prism coupler for intercepting an input light beam from a light source, the silicon-based prism coupler is permanently fixed on the SOI wafer such that a first surface of the prism coupler is substantially aligned with the flat surface of the SOI wafer Parallel and connected, the refractive index of the silicon-based prism coupler is equal to or greater than the refractive index of the silicon optical waveguide; A free-space micro-optical input element inserted between the light source and the silicon-based prism coupler to collimate, shape, and direct the beam to a specific entry point and angle of incidence of the silicon-based prism coupler; an evanescent wave coupling region disposed between the silicon-based prism coupler and the silicon optical waveguide; and Free-space micro-optics placed in the path of a beam of light exiting the output surface of a silicon-based prism coupler to shape, collimate, or converge the beam and direct the beam to a receiving element. 2. The optical coupler device of claim 1, wherein the device further comprises a light source coupled to the free-space micro-optical input element. 3. The optical coupling device of claim 2, wherein the wavelength of the light source falls within the range of 1.1-1.65 [mu]m. 4. The optical coupling device of claim 2, wherein the output beam of the light source is substantially single mode. 5. The optical coupling device of claim 2, wherein the full intensity of the light source falls substantially within ±5 nm of the central wavelength. 6. The optical coupling device of claim 2, wherein the light source is an edge emitting laser diode. 7. The optical coupling device of claim 5, wherein the micro-optical free-space input element following the edge-emitting laser diode comprises a first micro-optic element that reduces the divergence angle of the output beam perpendicular to the junction To the order of the divergence angle of the output beam parallel to the junction, the astigmatism is corrected and a circular beam is produced, and a second micro-optical element subsequently collimates the beam. 8. The optical coupling device of claim 6, wherein the micro-optical free-space input element following the edge-emitting laser diode comprises a gradient-index microcylindrical lens that collimates the output beam perpendicular to the junction, and a subsequently parallel The output beam of the diode junction is collimated by the second microcylindrical lens. 9. The optical coupling device of claim 6, wherein the micro-optical free-space input element after the edge-emitting laser diode comprises a first spherical lens to collimate the beam, followed by a converging beam between the diode and the silicon-based prism coupling A second spherical lens on the receiving fiber optic component between the receivers. 10. The optical coupling device of claim 6, wherein the micro-optical free-space input element behind the edge-emitting laser diode comprises a first aspheric mirror that collimates the beam, a second aspheric mirror that then converges the beam between the diode and the silicon on the receiving fiber optic component between the base prism coupler. 11. The optical coupling device of claim 6, wherein the micro-optical free-space input element behind the edge-emitting laser diode comprises a micro-optic waveplate for rotating the polarization direction. 12. The optical coupling device of claim 2, wherein the light source is a vertical cavity surface emitting laser diode. 13. The optical coupling device of claim 12, wherein the micro-optic free-space input element behind the vertical cavity surface emitting laser diode comprises a micro-optic collimating lens. 14. The optical coupling device of claim 13, wherein the micro-optical collimating lens is a silicon microlens. 15. The optical coupling device of claim 12, wherein the micro-optical free-space input element following the VCSEL comprises a micro-optic waveplate for rotating the polarization direction. 16. The optical coupling device of claim 12, wherein the micro-optical free-space input element following the VCSEL comprises an optical element that converts an incident light beam in an unknown polarization state into one having a defined Two independent output beams of known polarization state, and the second beam is separated from the first beam but kept substantially parallel. 17. The optical coupling device of claim 2, wherein the light source is an optical fiber. 18. The optical coupling device of claim 17, wherein the fiber is single-mode and supports arbitrary polarization states. 19. The optical coupling device of claim 17, wherein the optical fiber is a single-mode polarization maintaining optical fiber. 20. The optical coupling device of claim 17, wherein the micro-optic free-space input element behind the fiber comprises a micro-optic collimating lens. 21. The optical coupling device of claim 20, wherein the micro-optics collimating lens comprises a lens fused to the optical fiber to form a lensed optical fiber. 22. The optical coupling device of claim 21, wherein the diameter of the collimated beam emerging from the lensed optical fiber is in the range of 10-110 [mu]m. 23. The optical coupling device of claim 17, wherein the micro-optical free-space input element after the fiber comprises an optical element that converts an incident light beam in an unknown polarization state into two independent output beams having the same known polarization state light beam, and the second light beam is spaced apart from the first light beam but kept substantially parallel. 24. The optical coupling device of claim 1, wherein the micro-optical free-space input element includes a refracting wedge of high index material to deflect the incident beam. 25. The optical coupling device of claim 1, wherein the micro-optical free-space input element comprises a reflective element that can be moved and rotated by an electronically actuated mechanism to cause movement and angular deflection of the incident light beam. 26. The optical coupling device of claim 1, wherein the micro-optical free-space input element comprises a diffractive optical element that angularly deflects an incident light beam. 27. The optical coupling device of claim 1, wherein the thickness of the evanescent wave coupling region is substantially constant. 28. The optical coupling device of claim 1, wherein the evanescent wave coupling region has a wedge-shaped thickness. 29. The optical coupling device of claim 1, wherein the device further comprises a receiving optical element for receiving the output beam from the free-space micro-optical output element. 30. The optical coupling device of claim 29, wherein the receiving element is an optical fiber. 31. The optical coupling device of claim 29, wherein the receiving element is a lensed optical fiber. 32. The optical coupling device of claim 1, wherein the input and output micro-optical elements, and the input and output surfaces of the silicon-based prism coupler are covered with an anti-reflection coating. 33. An optical coupling device providing a path for signals into and out of a silicon optical waveguide formed in a surface layer of a silicon-on-insulator (SOI) wafer comprising a silicon A silicon optical waveguide layer on an insulating layer on a substrate, the optical coupling device comprising A silicon-based prism coupler permanently fixed on the SOI wafer such that the first surface of the prism coupler is substantially parallel to and connected to the straight surface of the SOI wafer, and the refractive index of the silicon-based prism coupler is equal to or greater than the refractive index of the silicon optical waveguide; optical elements forming an integrated component of the silicon-based prism coupler for collimating, shaping, and directing the input beam to a specific entry point and angle of incidence on the coupling surface of the silicon-based prism coupler; an evanescent wave coupling region disposed between the silicon-based prism coupler and the silicon-based optical waveguide; and A free-space micro-optics output element placed in the path of a beam exiting the output surface of a silicon-based prism coupler to shape, collimate or converge the beam and direct the beam to a receiving element. 34. The optical coupling device of claim 33, wherein the microlenses are formed in the surface of the silicon-based prism wafer instead of the connection surface of the SOI wafer to collimate the incident light beam. 35. The optical coupling device of claim 33, wherein the diffractive optical element is fabricated in the surface of the silicon-based prism wafer instead of the bonding surface of the SOI wafer to shape the incident beam or diverge or angle the incident beam. 36. The optical coupling device of claim 33, wherein the angled surface is anisotropically etched into the silicon-based prism coupler to angularly deflect the incident beam through internal reflection. 37. The optical coupling device of claim 33, wherein the sub-surfaces in the silicon-based prism coupler are covered with a thin metal layer that acts as a reflective element to deflect the angle of the incident light beam. 38. The optical coupling device of claim 33, wherein the evanescent wave is evanescent at the point where the light beam enters the waveguide from the silicon-based prism coupler at the input prism coupling surface and exits the waveguide into the silicon-based prism coupler at the output prism coupling surface. The coupling region is wedge-shaped such that the beam at all points in the optical coupling device outside the SOI wafer waveguide has an intensity distribution in a substantially Gaussian mode. 39. The optical coupling device of claim 38, wherein the substantially Gaussian mode intensity profile of the output beam obtained using the wedge-shaped evanescent coupling layer enables efficient coupling of the beam into the receiving optical fiber. 40. The optical coupling device of claim 33, wherein the thickness of the waveguide of the SOI wafer is selected such that light emitted from the light source parallel to the wafer surface and incident on the input prism face is received by the silicon-based prism coupler at an angle Refraction, this angle is related to high coupling efficiency for a specific wavelength. 41. The optical coupling device of claim 33, wherein the thickness of the waveguide of the SOI wafer is selected such that light emitted from the light source perpendicular to the wafer surface and incident on the input prism face is received by the silicon-based prism coupler at an angle Refraction, this angle is related to high coupling efficiency for a specific wavelength. 42. The optical coupling device of claim 33, wherein the device further comprises a light source. 43. The optical coupling device of claim 42, wherein the light source is a vertical cavity surface emitting laser diode having a suitable wavelength. 44. The optical coupling device of claim 33, wherein the thickness of the waveguide of a certain SOI wafer is selected so that the light emitted from the light source and incident on the input prism face is refracted by the silicon-based prism coupler, so that the light beam passes through the prism The projection on the coupling surface remains essentially constant over a wide range of wavelengths. 45. The optical coupling device of claim 1, wherein the thickness of the waveguide of a certain SOI wafer is selected such that, for the same wavelength, the thickness of the evanescent wave coupling region optimized for coupling efficiency for a given wavelength and input beam size is substantially is equal to a quarter wave thickness of the material making up the evanescent wave coupling region. 46. The optical coupling device of claim 45, wherein the evanescent wave coupling layer and the anti-reflection coating of the silicon-based prism wafer are simultaneously processed in one processing step.

Images